To Retail Out-of-Stock Reduction: A Comprehensive Guide

A Comprehensive Guide
To Retail
Out-of-Stock
Reduction
In the Fast-Moving
Consumer Goods Industry
A research study conducted by:

Thomas W. Gruen, Ph.D., University of Colorado at Colorado Springs, USA
And Dr. Daniel Corsten, IE Business School Madrid
A Comprehensive Guide
To Retail Out-of-Stock
Reduction In the Fast-Moving
Consumer Goods Industry
A research study conducted by:
Thomas W. Gruen, Ph.D., University of Colorado at Colorado Springs, USA
And Dr. Daniel Corsten, IE Business School Madrid
This study was funded by a grant from the Procter & Gamble Company
Acknowledgements
Supporting Trade Associations:

Grocery Manufacturers Association
Food Marketing Institute
National Association of Chain Drug Stores
Partner Academic Institutions

Technical Eindhoven University of Technology, TU/e Retail Operations Group, The Netherlands
MIT
Special thanks and recognition for his significant contributions to this report:
J.P. Brackman, Global Retail Presence Manager, Procter & Gamble Company
Special thanks to the data partners who contributed significantly to this report
© Copyright 2007 by the Grocery Manufacturers Association (GMA), Food Marketing Institute (FMI), National Association
of Chain Drug Stores (NACDS), The Procter & Gamble Company (P&G) or the University of Colorado at Colorado
Springs. All rights reserved. No part of this publication may be reprinted or reproduced in any way without express
consent from GMA, FMI, NACDS, P&G or the University of Colorado at Colorado Springs.
Contact for permissions: [email protected]; [email protected]
ISBN: 978-3-905613-04-9
Table of Contents
Executive Summary i-vi
Introduction............................................................................................................................................................ 1-6
Overview, Objectives, and Key Findings................................................................................................................................................... 1
What is an Out-of-Stock?.............................................................................................................................................................................1-3
Guide to Using this Report.........................................................................................................................................................................3-5
Chapter 1: Understanding the Total Cost of OOS............................................................................................... 7-12

1-1. Introduction and Overview With All of This Technology, Why Isn’t Availability Higher?................................................ 7
1-2. A Generally Unrecognized Problem: Costs of OOS > Lost Sales...........................................................................................7-9
1-3. Root Cause Analysis Drives a Distinction Between Store and Shelf OOS Conditions . ............................................ 9-10
1-4. We Know How Consumers Respond to OOS Conditions ..................................................................................................10-11
Chapter 2: Understanding OOS Through Measurement.................................................................................. 13-23

2-1. Understanding the Measurement and Identification of OOS..........................................................................................13-18
2-2. Focus on the Items that Cause the Majority of OOS: There Really Aren’t Very Many Fast Movers......................18-19
2-3. Viewing OOS Patterns.....................................................................................................................................................................20-23
Chapter 3: Lowering OOS Rates—Our Hypotheses and Related Studies....................................................... 25-40

Overview............................................................................................................................................................................................................25
3-1. Ordering What You Need When You Need It: The Case for Item Data Accuracy........................................................25-26
3-2. Ordering What You Need When You Need It: The Case for Inventory Accuracy........................................................27-30
3-3. Demand Forecasting Accuracy....................................................................................................................................................30-32
3-4. Having the Goods Arrive When You Need Them . ................................................................................................................32-34
3-5. Making Room on the Shelf: The Case for Demand-Based Planograms .......................................................................34-36
3-6. Knowing Where Everything Should Be: The Case for Planogram Compliance .........................................................36-38
3-7. Knowing Where Everything Should Be: Keeping the Shelves Straight.........................................................................38-39
Chapter 4: Getting and Keeping Lower Levels of OOS...................................................................................... 41-48

Overview.............................................................................................................................................................................................................41
4-1. The Ability to Measure Provides the Ability to Focus...........................................................................................................41-42
4-2. Taking Action......................................................................................................................................................................................42-44
4-3. A Comprehensive Approach to Reducing OOS Events, Duration, and Losses..................................................................44
4-4. Practical Steps to Address OOS of High Demand Items......................................................................................................44-46
4-5. Bringing It All Together and Looking Ahead...........................................................................................................................46-47
Appendix................................................................................................................................................................ 49-56
Authors’ Information......................................................................................................................................... 57-58
A Comprehensive Guide To Retail
Out-of-Stock Reduction In the
Fast-Moving Consumer Good Industry
Executive Summary
Overview Out of Stock Definition is Clarified
This research report provides a comprehensive This report catalogs the meaning of various “out of stock
examination of the foundational knowledge, measurement rates” that have been reported in previous studies. We hope
approaches, and strategies used to reduce retail out-of- to avoid further confusion caused by vague reporting of the
stocks (OOS) in the fast-moving consumer goods (FMCG) data collection and calculation methodology used to establish
industry. Its objective is to provide a guide to FMCG the rate. The three common types of measurement— (1)
retailers that seek to reduce their costs associated with OOS audit of physical inventory, (2) analysis of point of sale data
items, while simultaneously enhancing shopper satisfaction (POS), and (3) use of perpetual inventory data (PI)—each
with sustained lower levels of OOS. As outlined in Exhibit measure different aspects of OOS, and thus report different
A, this research report: OOS rates. Further ambiguity arises from differences in
1. defines retail OOS metrics to resolve the confusion what is counted – (a) instances, (b) units sales losses, or
surrounding previously reported “OOS rates” and shows (c) monetary sales losses. Generally, audits count instances
the total costs of OOS, beyond lost sales (Chapter 1); observed at a point in time (when the audit took place).
2. compares the three basic approaches to the measurement PI data is typically used to count instances over some time
of OOS, illustrates how OOS measurement can be interval (perhaps a week). POS data analysis can yield a
directly linked to root causes, and shows where most broader set of measurements including each of (a), (b), and
OOS sales losses occur (Chapter 2); (c) above.
3. systematically examines seven causes of OOS, from
forecasting to merchandising, showing the impact on We make a clear delineation between an OOS event
overall OOS levels of addressing each (Chapter 3); and (an instance of an item being unavailable for sale as
4. provides a flexible approach to reduce OOS that intended), and the attributes of the OOS event, and
retailers can easily adopt, can be effective even with low statistical descriptions of collections of OOS events
initial level of resource commitment and can scale with (expressed as an OOS rate). These attributes include:
increased resource commitment (Chapter 4). 1) number of occurrences over time, 2) number of
simultaneous occurrences, 3) duration, 4) shelf availability,
Our extensive research shows that retailers can 5) lost unit sales, 6) lost monetary sales, and 7) number of
sustain OOS reductions below the industry average of customers impacted.
8.3 percent. The core requirement is the development of
an effective measurement system – one that is accessible, Store vs. Shelf OOS Perspective is Established
timely, inexpensive, reproducible, and generally undistorted. One of the keys to efficiently reducing OOS is a clear
Without this core ability, it is impossible to determine delineation between types and their distinct underlying
progress, assign responsibility for tasks, and maintain causes. From the retailer perspective, the three main OOS
accountability for results. When the measurement stops, types are the distribution center OOS, store OOS, and shelf
people go back to their old way of operating, and the OOS OOS. While much of the industry is focused on distribution
return to their previous levels. OOS, this report focuses on store and shelf OOS types.
Gruen & Corsten 2008

Executive Summary
Exhibit A: Overview of the Report
Steps of Sustained Lower Out-Of-Stocks

1) Motivation: Calculating “The Size of the Prize”
2) Measurement: Knowing Where to Focus
3a) Lowering OOS: 3b) Lowering OOS:

Store-Based Ordering Shelf-Based (Operations)
Forecast & Replenish-

Data Order ment Shelf Shelf Shelf
Accuracy Accuracy • Delivery Capacity Implemen- Management
• Product • Understated Frequency • Case tation • Replenish-
Data or • Cycles Packout • POG ment
Accuracy Overstated • Fill Rate • Time Supply Compliance • On-Shelf
• Inventory Dueto OOS • Execution • Demand Execution
Accuracy in POS Data Based POGs
• POS • Computer
Data Ordering
Accuracy • Manual
Ordering
4a) Approach to Lowering OOS: Assessment and Implementation
4b) Sustaining Lower OOS
A store OOS occurs when the store is completely out The Total Costs of Out-of-Stocks are Examined
of inventory. Excessive store OOS arise from mistakes in The impact of OOS extends well beyond the lost sales
ordering, demand forecasting, or supply chain. Shelf OOS of the OOS item alone. A variety of strategic and operational
occur when there is inventory in the store, but the item is costs apply to both retailers and suppliers including
not on the shelf. The root causes of excessive shelf OOS decreases in store and brand equity and attenuated impact
are usually store processes, especially shelf space allocation,
of promotions and trade promotion funds. OOS creates a
restock frequency, and ongoing monitoring of shelf stock
ripple effect by distorting demand and leading to inaccurate
for promoted items. Thus store and shelf OOS each have
forecasts. Retailer costs also include the time employees
a different set of associated solutions. Understanding the
difference and specifically addressing the root causes of each spend trying to satisfy shoppers who ask about a specific
type will yield major reductions in overall OOS rates because OOS item. For a typical U.S. grocery store, the cost amounts
most OOS events are at the retail shelf or the store. In to $800 per week. The corollary for shoppers is the amount
other words, the distribution center could have supplied, or of time spent waiting for resolution that could be spent more
did supply the item – but there was a problem in retail store productively for the retailer in shopping—an estimated 20
processes or execution. percent of the average time for a shopping trip.
ii A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry
Executive Summary
The Relationship of Volume on Out of Stock Rates Measurement Must Point to the Root Cause
In this study we examine the relationship between sales Regardless of the measurement system used to track
velocity and OOS sales losses. Not surprisingly, fast moving OOS— manual audit, POS data estimation, or perpetual
items incur proportionally greater OOS sales losses, but the inventory—it must be sustained, and it must point towards
degree and concordant lost sales were surprising. As shown root causes. Due to their high expense and difficulty to
in Exhibit B, the classic 80/20 rule applies for item sales scale, manual audits are usually not sustainable, and they do
(20 percent of items comprise 80 percent of the total store not provide a measure of sales loss. However, they can be
sales in an average week). On the busiest day, 65 percent or effective when targeted at the most crucial products (either
more of the items did not sell at all, and on an average day, high velocity items or strategically important items such as
75 percent or more of the items do not sell at all. Regarding “never outs” or preferred private brands), and when a second
level of analysis is incorporated that links each OOS event
Exhibit B: Daily, Weekly, and Annual Item Unit Sales to its likely root cause. A systematic means of assigning each
identified OOS event to a set of pre-determined root causes
can be implemented at a relatively low initial cost. However,
it is costly to scale to a large number of items.
The use of point of sale (POS) data is a viable

measurement method for many store formats. There are a
number of companies that have developed algorithms to
estimate OOS from POS data, and some retailers have
developed their own in-house systems. POS measurement
systems can be sustained, scaled and are able to deliver sales
loss and duration measures. The accuracy of estimating OOS
using POS data is 85 percent or greater, which is equivalent
or greater to the accuracy of manual audits (where human
Credit: Standard Analytics, 2006 error is present). One recent development of using POS
data calculation is the ability
to discern visible patterns in
Exhibit C: OOS Patterns Indicating Inadequate Replenishment Schedule out of stocks and thereby point
directly at possible root causes
and potential solutions—all done
electronically. Exhibit C shows
one of several patterns that have
been identified as typical out of
stock patterns. The pattern here
shows that the OOS period
occurs nearly the same week
after week. Most likely, there are
two deliveries a week, but four
deliveries are needed. Alternative,
the store could increase the
safety-stock of this item. This
approach shows enormous
Credit: Standard Analytics
promise in reducing OOS due
to its intuitive nature.
the relationship of the sales volume to out of stocks, the A third approach to measurement, perpetual inventory
high demand items had almost six times the levels of lost (PI) measurement systems can also be sustained, scaled
sales from the out of stocks they encountered than did low and deliver sales loss and duration measures. However, PI
demand items. These findings provide clear evidence that systems suffer from the lack of on-hand accuracy necessary
focusing on a relatively small number of SKUs can be an to make them consistently good measures. Algorithmic
effective strategy to lower OOS sales losses. approaches to estimating and improving on hand inaccuracy
Gruen & Corsten 2008 iii

Executive Summary
are being developed and implemented, and a best practice 3. Demand Forecasting Accuracy. Ideally a demand forecast
approach of manual techniques for improving PI accuracy is should be the same as a sales forecast, however they
provided in this report. invariably differ, largely because of the impact of sales
variances caused by OOS. Whenever a shopper does
Root Causes and Solutions not buy or shifts their buying pattern due to an OOS, it
Moving from left to right across the trapezoids shown adjusts the demand history away from the sales history
in Exhibit A, we researched seven key, different root causes and no one can see the true demand history. Merging
and solution areas. POS lost sales history with the sales history can more
1. Product Item Data Accuracy. Product data inaccuracy closely represent true demand and lead to better demand
creates an unstable foundation for ordering and forecasts. When we further examined the impact of
forecasting. Commonly referred to as “data synch,” individual store managers adjusting merchandising
there are clear impacts on out of stocks when product quantities from suggested computer assisted ordering
data issues are excessive. The primary recommendation (CAO) quantities, we found that store personnel
focuses on collaborative synchronization of data between underperform even imperfect CAO demand forecasts.
suppliers and retailers using a third party vendor. 4. Store and Shelf Replenishment. Using the POS
We also show how the use of a parent-child product measurements we were able to identify patterns that
relationship system can enhance product data accuracy. showed when store replenishment (from the distribution
2. Ordering and Inventory Accuracy. We identified a center or by DSD vendors) was too infrequent. We
variety of store issues that create PI system inaccuracy also found a positive relationship between backroom
(especially on hands). The level of PI inaccuracy was inventory and OOS, and thus recommend matching
stunning, as PI accuracy (where the PI exactly matched delivery schedules to meet the demand on the
the on-hands) ranged from 32 percent to 45 percent in shelf, rather than maintaining backstock (except for
the four studies we conducted or examined. Exhibit D promotional and other specific items).
shows the distribution of PI accuracy for the best case 5. Shelf Space Allocation. We found that 91 percent of
we encountered. Phantom inventory (when PI system the SKUs are allocated shelf space based on case pack
on-hand is greater than true physical product on- size, and that 86 percent of the inventory on shelves is
hand) is a major cause of OOS, particularly store OOS, in excess of seven days supply. Given that the shelves
because the reorder system does not recognize how are crowded, and that the fast moving items have six
low store inventory levels are. For the retailer shown in times the lost sales due to OOS than their slower
Exhibit D, items with correct on-hands had OOS event moving counterparts, there is a strong case to be made to
rates of 4.1 percent and had a rate of 8.9 percent where reallocate additional shelf space to the small number of
on-hands were not accurate. faster moving items using demand-based planograms.
6. Planogram Compliance. We found that a 10 percent
change in planogram compliance resulted in a 1 percent
Exhibit D: Audited PI Accuracy of U.S. Retail Chain change in the level of OOS. Thus, categories that have
high planogram compliance levels (90 percent or better),
this would be a low priority. For retailers with low
compliance levels, addressing planogram compliance can
be a good way to lower OOS. The report also provides
a best practice methodology for measuring planogram
compliance.
7. Item Management. We examined the stocking practices
that would affect manual ordering systems. We focused
on three well-known links to OOS: 1) covering holes, 2)
hiding product, and 3) shelf-tagging accuracy. In a new
study we found that simple adherence to these practices
had a huge effect on out of stocks, reducing OOS levels
by about 40 percent, as shown in Exhibit E. While most
retailers have policies for these practices, many were not
enforced.
iv A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry
Executive Summary
Exhibit E: Impact of Disciplined Shelf Management Practices on OOS Sales Losses level of lost sales due to OOS (or which items
have been strategically identified on other
criteria to have higher availability levels), and
which stores have the highest level of lost sales
due to OOS.
2. Define the targeted amount of OOS
reduction desired, and then allocate the
amount of gain to be achieved from product-
based OOS reductions (store or ordering
types of solutions) and from store-based
OOS reductions (shelf or operations types of
solutions). For example if a goal of $500,000
lost sales reduction has been established,
determine how much of the goal will be
obtained from ordering type solutions such
as PI data accuracy and how much should
Credit: Data Ventures be gained from store type solutions (such as
demand-based planograms). In this example,
it may mean focusing on the top 300 products
RFID Technology and Shelf Out of Stocks and the worst 24 stores to get them to their target rate.
Due to technological and financial reasons, most radio The point of the recommendation is that this approach
frequency identification (RFID) applications have been can be implemented by any retailer, and the degree
limited to tags on pallets and cases and have not descended of implementation can vary based on the available
to the individual item level, where RFID shows great resources
promise to address shelf OOS. However, at the case and 3. Apply the identified solutions in the assessment
pallet level, RFID applications can track when the cases specifically to those products (across all stores) and those
are delivered to the store’s backroom, and when they move stores (across all products). This will provide an estimate
from the backroom to the store floor and vice versa. As a of the resource level needed to achieve the goal, and the
result, RFID has been shown to reduce shelf OOS for high degree to which adjustments can be made based upon
velocity items that require that the store hold large levels of the available resource. This targeting will keep work
backstock. RFID applications can enhance sorting of cases and disruption to a minimum. Once the solutions are
coming off a delivery truck. Items that are known OOS get implemented across these products and stores, a greater
identified quickly for immediate stocking, while items that level of resources could be applied to additional products
are still available to the shopper but have room on the shelf and stores.
for a full case get secondary attention. Cases that are back-
stock remain in the backroom, rather than being taken to There are a few important things to keep in mind with
the sales floor and returned. RFID requires disciplined shelf this methodology. First, both products and stores should be
stocking practices. A case that cannot be completely stocked addressed, not exclusively one or the other. Working in one
on the shelf becomes a problem when returned partially full provides synergy to the other area, and increases the overall
because the RFID does not recognize a partial case in the solution effectiveness. Second, some of the product gains
backroom. In addition, RFID is being effectively applied will come in the targeted stores, thus are not fully additive in
to recognize shrink at the case level, where the impact of their impact. Finally ongoing and permanent measurement
unrecognized shrink can have a large effect on OOS due to should be built into the process in order to sustain the gains
its large impact on inventory inaccuracy. in OOS reduction.
A Methodology to Get and Keep Low Levels of Out Final Conclusions

of Stocks 1. Out of stocks CAN be reduced on a sustained basis.
A simple three-step approach can be applied to address 2. Measurement lays the foundation to be able to focus
OOS. This approach is completely flexible based the amount resources where losses are the most critical.
of resources that can be devoted to addressing OOS and the 3. Selecting and maintaining a focus is crucial for success.
desired amount of OOS reduction. 4. There are proven solutions for many identified root
1. When assessing OOS, create both a product ranking causes.
and a store ranking, i.e., which products have the highest 5. There is a simple, yet workable plan to achieve results.

vi A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry
Introduction
Overview and Objectives After spending three years studying approaches to

In 2002 we published a study on retail out-of-stock lowering OOS levels, our collective thinking led us to
(OOS) levels of fast-moving consumer goods (FMCG) organize the general approach to addressing OOS that is
products that included many key findings including that the shown in Figure 1. Contained in this relatively simple set
overall level of OOS in developed countries world-wide to of steps is a series of hypotheses and studies that test each
be 8 percent, that 75 percent of the cause was due to retail hypothesis, reading and assimilating study after study of
store practices (opposed to up-stream supply issues), and others who have been doing research in this field, meeting
that on average 30 percent of the items that were OOS were after meeting with expert consultants, suppliers and retail
purchased at another retail store. The implications of our personnel, hours and hours on the sales floors and back
study suggested that retailers on average lose 4 percent of rooms of countless retailers in the USA and Europe, and
their annual sales due to OOS items, and that lost sales due enough air miles to keep the authors platinum.
to OOS items on average cost manufacturers $23 million for
every $1 billion in sales. A Few Key Findings
So what did we learn from all of our efforts? Here are a
These findings have been distributed, recognized few findings:
and published widely, and they have become the basis for • New retail support technologies, such as RFID, have
numerous academic and industry research studies, as well changed our thinking about OOS as a single problem,
as for many business projects that have addressed on-shelf to separating OOS examination to “not in the store” and
availability. These various studies and projects have focused “not on the shelf.”
on new ways of measuring OOS, identifying and analyzing • An examination of SKU sales frequency shows that only
root causes, and examining ways to increase on-shelf a small portion of items make the big difference. This
availability. Individually, each of these studies and projects has implications for identification, ordering, and shelf-
enlightens various aspects related to OOS, and there is now space allocation.
a need to bring these issues together and provide the FMCG • There are new methods of measuring OOS that don’t
industry a guide for collectively addressing OOS problems take much labor, and need to be implemented.
and systemically increasing retailers’ ability to offer shoppers • Regardless of the method used, measurement and
increased levels of availability. identification of OOS can point directly to the root
cause, and thus the solutions become obvious.
That is the purpose of this report. In it, we share the • Accuracy of data—product data, inventory, and point of
results of several studies we have undertaken since 2002, sales data—is the foundation for keeping product in the
as well as bring together the best knowledge generated by store and on the shelf.
many others who have been addressing OOS issues and
who have agreed to partner with us in this endeavor. In What are Out-Of-Stocks?
our 2002 report, we provided baselines and benchmarks Previous studies of OOS have used a variety of
for understanding and measuring the extent, consumer definitions, and this has resulted in confusion regarding what
responses, and root causes of OOS. In this report, we is being measured and what an “OOS rate” actually means.
bring the discussion full circle, providing baselines and We have identified several related measurements, and there
benchmarks for measurement and identification of OOS, as are two fundamental concepts that need to be understood:
well as providing a guide to those responsible for addressing
OOS — those organizations interested in increasing and Concept 1: The “OOS event” refers to what an “out-of-
sustaining higher levels of on-shelf availability. stock” is (i.e., how we know one when we see one). An
OOS event occurs when, for some contiguous time, an
item is not available for sale as intended. If the retailer

Introduction
intends an item to be for sale, but there is no physical previous example, the shelf availability rate for the
presence of a salable unit on the shelf, then the item is item would be 88 percent, meaning that the item was
deemed to be OOS. The OOS event begins when the available for sale during 88 percent of the time the
final saleable unit of a SKU is removed from the shelf, store was open.
and it ends when the presence of a saleable unit on the
shelf is replenished. 5. Lost Sales: While understanding and measuring
OOS occurrences and duration is important for
Concept 2: The attributes of the OOS condition refer assessing supply chain and store merchandising
to aspects of the OOS event(s) that can be measured effectiveness, the most important attribute for providing
and that can be calculated as an OOS rate. There are diagnostic direction for OOS attenuation strategies is the
multiple attributes that describe OOS events, and thus lost sales caused by an OOS. A lost sale occurs each
there are multiple OOS rates that are calculated and time a shopper who wants to buy an item that is listed
reported. Each attribute can be expressed as a rate over a for sale in the retail store cannot find the item in its
given measurement period. OOS attributes include: expected place and thus cannot purchase the item.
If the item tends to be a fast mover, there will likely
1. Number of occurrences over time: “Item OOS event be multiple lost sales during a single OOS event. If
rate.” This is typically measured as the simple number the item is a slower mover, there may be no lost sales
of OOS events for an item over a given unit of time, that occur during the OOS event, especially if it is a
for example, “six times per month.” This measurement short duration. There are two primary measures of lost
is useful when comparing the rate against benchmarks sales, both the units and the sales monetary volume.
or among items in a category. A related measure is the number of customers that are
impacted.
2. Number of simultaneous occurrences: “Category 5a. Lost Unit Sales: “OOS lost unit sales rate.” Over
OOS event rate.” For this measure, the number of a given period of measurement, the OOS lost
items in the category that are OOS at the time the unit sales rate is calculated as the total number of
measurement is taken are summed and expressed as estimated sales unit losses due to OOS divided
a percentage of the total number of items intended by the total number of sales units sold plus the
for sale. For example, if the category has 50 items estimated sales unit losses.
intended for sale, and at a given time there are five 5b. Lost Monetary Sales: “OOS sales loss rate.” Over
items that are OOS, then the Category OOS event rate a given period of measurement, the OOS sales
is 10 percent. This is the most commonly reported rate loss rate is calculated as the total estimated sales
because it is easy to calculate when taking a manual volume (dollars, Euros, or other monetary unit)
audit of the category. losses due to OOS divided by the total sales
plus the estimated dollar sales losses. Derivative
3. Duration: “OOS duration rate.” Over a given period loss rates can be calculated using the relevant
of measurement, the OOS duration rate is calculated measure, such as gross margin. However, most
as the total time that the item is OOS divided by the loss rates are calculated on gross sales volume.
total selling time available of the measurement period. Unless specifically specified, lost monetary
For example, if a store is open 24/7, and if during a sales calculations do not consider the impact of
given week of measurement an item was OOS for 20 shoppers switching to other products. Thus the
hours, then the OOS duration rate would be 12 percent actual monetary loss to the retailers will be less
(20 OOS hours/168 possible selling hours). It is than the summed estimates of the OOS of each
important to note that the OOS duration over a given individual item.
period may (and typically does) consist of multiple
OOS events, and duration of each event is summed 6. Number of Customers Impacted: “OOS Customer
over the measurement period. The OOS duration rate Impact Rate.” This is measured as the number of
measures the lost selling time for the item. shopping baskets that an item would have appeared
had the item been always available throughout the
4. Shelf Availability: “Shelf Availability Rate.” This is measurement. It is mathematically determined as: 1
directly related to the OOS Duration Rate, which – [(the number of estimated baskets the item would
refers to the probability that shoppers will find the have appeared - the actual number of baskets the
item when they enter the store. This is calculated item appeared) / the number of estimated baskets the
as 100 percent - OOS Duration Rate. Thus, for the item would have appeared]. For example, if during
A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry
Introduction
a typical week an item—if always available—would Guide to Using this Report

be estimated to appear in 100 shoppers’ baskets, but This report is based on insights and solutions that will
it actually appeared in 95 baskets due to the item assist managers in decision making and resource allocation
being OOS, then the customer impact rate would be to improve on-shelf availability. The report is organized into
5 percent. The customer impact rate can theoretically four chapters, following the steps in Figure 1.
be equal to the OOS lost unit sales rate, but it is
normally lower, as it accounts for multiple unit sales Chapter 1: Understanding the total cost of OOS
per customer. This chapter makes the business case for addressing
OOS from both an operational and a strategic level.
Summary of OOS Rates Discussion of the rationale, conclusions and summary of
1. Item OOS Event Rate: solutions at a high level are included in the body of the
The number of OOS events for an item over a given unit report, while detailed descriptions of the solutions appear
of time in the appendices. This chapter starts where our previous
2. Category OOS Event Rate: study left off. It takes the basic cost of lost sales due to
The number of items in the category that are OOS at OOS that was determined in our previous study, and it goes
the time the measurement expressed as a percentage of forward to show how the complexity of the effects of OOS
the total number of items intended for sale results in other previously undocumented cost areas that are
3. OOS Duration Rate: incremental to the immediate costs incurred by lost sales. For
During a given measurement period, the total time that example, employee time costs, shopper inconvenience costs
the item is OOS / the total selling time available and continued inaccurate ordering by retailers who must
4. Shelf Availability Rate: forecast in the presence of unknown demand. One focus
100 percent - OOS Duration Rate is on a recent distinction in addressing OOS, separating
5. OOS Lost Unit Sales Rate: store vs. shelf issues, a distinction becoming increasingly
The total sales unit losses due to OOS / (the total emphasized by those using RFID to address OOS.
number of sales units sold + the total sales unit losses)
6. OOS Sales Loss Rate: Chapter 2: Understanding OOS
The total monetary sales volume losses due to OOS / through measurement
(the total sales + the estimated dollar sales losses) To know where one wants to go (lower OOS), one
7. OOS Customer Impact Rate: needs to know where they are. In this study we compare
1 – [(the number of estimated baskets the item would the primary ways that OOS are measured, and we show
have appeared - the actual number of baskets the item how measurement needs to point towards root causes and
appeared) / the number of estimated baskets the item solutions. New light in understanding OOS is revealed
would have appeared] through aggregate item movement in the store, where the
analysis shows that a relative small number of items can
Key Distinction of OOS Events be considered “fast movers.” Once attention is focused on
these items, the chapter moves to examine different ways
to measure OOS—both electronic and manual. Other
Shelf OOS: Describes OOS events where a salable
attributes such as duration and frequency are addressed, as
item is physically somewhere in the store, but not readily
various measurement methods explore the attributes. New
available to the shopper because it is somehow hidden
attributes of OOS are presented, including OOS patterns
from the shopper in its expected place; i.e., it is not found
that can be identified. In all types of measurement, the
in the back room, it is in the wrong location, it is hidden
emphasis is on using measurement to identify root causes,
behind other products or otherwise. Store OOS: Describes
which can then be applied to solutions to reduce OOS.
OOS events where the item is not physically in the store,
as a result of forecasting or ordering inaccuracy, ordering
practices and delivery processes.

Introduction
Figure 1: Overview of the Chapters of This Report
Steps of Sustained Lower Out-Of-Stocks

1) Motivation: Calculating “The Size of the Prize”
2) Measurement: Knowing Where to Focus
3a) Lowering OOS: 3b) Lowering OOS:

Store-Based Ordering Shelf-Based (Operations)
Forecast & Replenish-

Data Order ment Shelf Shelf Shelf
Accuracy Accuracy • Delivery Capacity Implemen- Management
• Product • Understated Frequency • Case tation • Replenish-
Data or • Cycles Packout • POG ment
Accuracy Overstated • Fill Rate • Time Supply Compliance • On-Shelf
• Inventory Dueto OOS • Execution • Demand Execution
Accuracy in POS Data Based POGs
• POS • Computer
Data Ordering
Accuracy • Manual
Ordering
4a) Approach to Lowering OOS: Assessment and Implementation
4b) Sustaining Lower OOS
Chapter 3: Lowering OOS rates— perpetual inventory accuracy and planogram compliance,
our hypotheses and related studies we also share best-practice methods that were developed for
Building on the enhanced understanding of OOS as each study.
presented in Chapters 1 and 2, Chapter 3 systematically
examines efforts to address OOS starting with the Chapter 4: Assessment and implementation to get
initial steps of accurate data all the way through to daily and keep lower levels of OOS
replenishment practices. We delineate a series of hypotheses While each of the components to addressing OOS
and demonstrate the proposed effect on OOS levels, levels have been addressed systematically in the two previous
providing the available supporting evidence through our own chapters, this chapter provides a systematic approach for
studies as well as from others’ studies. For two of the studies, managers to address their OOS levels. This serves as a guide
Introduction
where managers can methodically diagnose and assess their

most serious OOS malady, and then apply the appropriate
protocol to address it.
This chapter concludes with a vision for what a
retailer could realize when a culture of continuous quality
improvement is merged with the various technologies to
provide a seamless, accurate, informational base that will
assure that the products shoppers want are available when
they want to buy them.

RFID Features
Radio frequency identification (RFID) has been viewed
as a technology that can support and enhance efforts to
reduce OOS. With the use of electronic product codes
(EPC) on palettes and case-packs, RFID readers can greatly
enhance the ability to track products as they pass through
designated thresholds. Special call-out boxes in this report
feature the impact RFID can have in the quest to increase
on-shelf availability.

Introduction
Understanding the
Total Cost of OOS 1
1-1. Introduction: With All This New Technology, more complex promotions, and increased SKU proliferation.
In sum, there are a variety of reasons, but we have identified
Why Isn’t Availability Higher? the primary reasons:
In 2002 we published a study, Retail Out-of-Stocks: A • Demand forecasts are made with incomplete
Worldwide Examination of Extent, Causes, and Consumer information, and thus often under-estimate demand;
Responses, that established the worldwide average level of OOS • Inaccurate data from inventory systems provide
in retail in the FMCG industry to be about 8 percent. This incorrect ordering information;
report clearly showed the industry that the problem of OOS • Traditional retail practices such as using only case-pack
items was caused primarily by retail practices, and estimated size to determine shelf allocation (86 percent of the
that OOS was costing the industry billions every year. dollar inventory on the shelf represents more than 7 days
of supply) prevail, choking shelf space from the relative
Is 8 percent good or bad? It’s bad. From a shopper few fast movers, without consideration of time of supply;
perspective, this means that for every 13 items one wants • Item/SKU (stock keeping unit) proliferation—suppliers
to buy, one will be out of stock. From a management battle for shelf space by introducing “me-too” items,
perspective, our 2002 study showed that OOS cost retailers and are constantly changing UPC / GTIN (universal
4 percent of sales, and this translates to a similar 4 percent product code / global trade identification number)
reduction in the average retailer’s earnings per share. What information and thereby contribute to inventory
was also interesting about the 8 percent figure was also that database inaccuracy;
it had not changed from a major study published seven years • Promotional proliferation, generally at the urging of
earlier (Coca-Cola Research Council 1996). In fact nearly suppliers;
every report among the hundreds we have read and reviewed • Consolidation among retailers that bridge information
that deal with retail out-of-stock levels in the FMCG systems containing inaccurate legacy data;
industry continue to converge on that 8 percent number, • Pressure to reduce personnel cost resulting in inadequate
as if there was some DNA in the industry that predisposes labor supply.
retailers to this level of service.
We thought those findings would make retailers With so many drivers of OOS levels, managers need
wake up to the fact that one of the easiest ways to improve to know where to begin addressing the component that will
earnings would be to simply get and keep goods on the have the most impact. And in order to know this, we have to
shelf. And it has. Over the past three years, we have read have a better understanding of OOS and how they affect the
or reviewed studies conducted by many retailers who seek business of both suppliers and retailers.
to reduce OOS levels, and many have been successful at
reducing OOS levels. Moreover, retailers have been able to 1-2. A Generally Unrecognized Problem: The Costs of
use new technologies that incorporate point of sale (POS) OOS > Lost Sales
data to better understand demand. Inventory systems have It has become abundantly clear to us over the years
been implemented to keep better track of inventory in the that we have been studying OOS, that the direct sales loss,
store, on order, and in transit. RFID technologies are now which we estimated to be up to 4 percent—substantial as it
being used on pallets and cases which help retailers better is for both retailers and suppliers—is only one part of the
know what they have in the store. expense OOS items produce. Figure 2 provides an overview
of several additional effects of OOS. The total costs are both
With all of this attention and technology thrown operational and strategic, and these affect both suppliers and
at increasing retail product availability, why do shoppers retailers.
still face unacceptable OOS levels? Overall, technology • From a services delivery perspective, an OOS item
improvements have been offset by process complexity, such as indicates that a number of service failures have occurred,

Chapter 1
and these service failures point to lowered customer Large Small

satisfaction, decreased store and brand loyalty and Volume Volume
increased shopper costs. Example Example
• From an operations and supply chain management
Avg $ volume/week (000) 298 60
perspective, OOS distort inventory information that is
required for ordering and replenishment of the store and Avg $ transaction $27.34 $15.00
shelf. In addition, treatment of OOS items requires extra Avg # customers/week 10.900 4,000
process steps that could be avoided if systems were in
Shoppers encountering 1 or more OOS 40.0% 40.0%
place that would eliminate OOS.
• From a marketing and sales forecasting perspective, # shoppers encountering OOS 4,360 1,600
the presence of OOS items distorts the baseline on % times shoppers involve store labor 10% 10%
which demand forecasts are made. Since true demand Avg minutes spent by store employee 6 4
is unknown due to OOS items, some items are under-
Avg. Hourly wage rate $18.00 $18.00
forecasted, while other items are over-forecasted.
Avg Cost/week/store $785 $192
Estimate of Aggregate Personnel # stores in chain 100 100
Lost Time Cost Due to OOS Items Weekly cost labor in chain $78,478 $19,200
The following spreadsheet calculator shows how to Annual cost labor in chain $4,080,878 $998,400
make an estimate of the actual time spent on tracing OOS
by personnel. This only looks at the cost of looking for average weekly sales are approximately 20 percent of that of
items when asked by a customer. While OOS does not the large volume format, the average transaction size is slightly
directly drive additional cost of this labor, this labor could be more than half the large volume size, and that the employee
deployed to productive efforts of the store. only spends four minutes on average per customer request.
Figure 2 In the large volume store example, the weekly cost per
store is about $800 per week,
Manufacturers Retailers and for the smaller volume store,
Operational • OOS lowers the potential impact of promotions and • OOS distorts true shopper demand thus decreases the cost per store is about $200
trade promotion funds; forecasting and ordering accuracy; per week. When these figures
• OOS distorts true store demand, thus category • Operational costs are increased through personnel are annualized across all stores
management and related efforts are less accurate looking for OOS items in back room, providing “rain in a chain, the total costs are
and effective; checks” to shoppers, unplanned restocking, etc. substantial.
• OOS increases overall costs of the relationship with (Note: an example of how personnel costs might be
the retailer (increased post-audit activity, irregular estimated is provided below) In both the large and
ordering) small volume examples, these
conservative estimates quantify a
Strategic • Direct loss of brand loyalty and brand equity; • Direct loss of store loyalty;
typically non-documented cost
• OOS encourages trial of competitor brands; • Decreased customer satisfaction;
• Lowered overall effectiveness of Sales Team • OOS encourages shopping at competitor stores;
caused by out-of-stock items,
resources • Permanent shopper loss to competitors (shopper
where they redirect scarce store
switch rate is still undocumented, but annual cost is
labor away from productive
US $1 million per every 200 shoppers) activities. Retailers can construct
this simple spreadsheet to match
their specific situation
The following examples, for the large volume store, we use
the Food Marketing Institute (FMI) average transaction size OOS and Increased Shopper Costs
(US $) of $27.34, estimate that 40 percent of the shoppers This study has not made any direct measures of shopper
will encounter at least one OOS (consistent with a 10 costs, but OOS events clearly increase the total shopping
item list), conservatively assume that only one of every 10 trip cost to the shopper. The measurement of the aggregate
will contact an employee about the OOS, and the average shopper costs in terms of increased transaction costs, lost
wage and benefits cost for the employee is $18.00 hourly. time, increased decision making requirements, and a host of
The average weekly sales volume for supermarkets (FMI) other social and psychological costs (for example the lower
is slightly under $300,000, and we estimate that the store confidence of having to use an untested substitute) has never
employee will spend six minutes on average searching for the been calculated—nor are the total effects understood. If we
requested item. In the small volume store example, the consider the “flip side” of the examples shown in the previous
Understanding the Total Cost of OOS
section, in a single large volume grocery store OOS add Figure 3: Root Causes of OOS
some amount of increased shopping costs to more than 4,000
shoppers weekly, or more than 200,000 shoppers annually.
The costs to each shopper vary. The shopper who makes Total Store
a quick decision to substitute a similar size and priced item upstream ordering
in a category will incur low time and costs, and depending causes 28% and forecasting
on the item, it can have low psychological costs as well. 47%
Alternatively, the shopper who is pressed for time may
purchase a more expensive substitute if the psychological In the store,
substitution costs are high. A customer with high not on the
substitution costs will go to another store to get the product, shelves 25%
and this is a very expensive proposition for the shopper.
Shoppers who ask store personnel to locate an item Credit: Gruen, Corsten, and Bharadwaj 2002
waste time in their shopping trip, having to find personnel
as well as waiting to see if the item is available. With the
OOS Causes Worldwide Averages
average shopping trip being 28 minutes, a six-minute wait
Subsequent studies have confirmed our findings that a
represents more than 20 percent of the entire shopping
substantial number of products that are actually in the store
trip time, precious minutes that the shopper might spend
are not found on their intended shelf. There are multiple
examining a new item or attending to other merchandising primary causes for shelf OOS. These include:
efforts of the retailer. In short, while the shopper is actually • a breakdown of in-store processes that are designed
in the store, the presence of an OOS forces the shopper to to move back-stock (excess inventory kept in the back
allocate minutes to activities other than those that retail room of the store to cover for periods where sales are
management would most prefer them to engage. expected to be greater than the amount of shelf space
allocated to the item for the regular delivery cycle) to the
In an age of increased consumer understanding of shelf;
the cost of time, and the availability of the internet as an • labor availability (not enough store stocking labor
alternative channel, retailers need to consider ways to make available);
shopping as convenient as possible (or at least eliminate • labor training—store stockers do not have a system for
the unnecessary inconveniences). Shoppers not only have getting OOS “holes” filled first;
alternative stores for shopping, but now also have entirely • the “hole” on the shelf is filled with another item, thus
new alternative shopping channels. not identified as OOS;
• a lack of ability to know when restocking the shelf is
1-3. Root Cause Analysis Drives a Distinction of Store required;
vs. Shelf OOS Conditions • item is located in multiple locations in the store and is
As we have been examining OOS over the past several out in one area while still available in another location;
years, a new focus on approaching OOS on two levels and
has begun to emerge: separating issues pertaining to store • too much product in the backroom—previous studies
OOS as opposed to shelf OOS. One can attribute the show a positive correlation of backroom inventory and
causes of store OOS to forecasting, ordering, delivery and shelf OOS.
upstream supply problems. In our 2002 study, there was little
information available on the impact of inventory inaccuracy This situation is particularly frustrating since the order
in perpetual inventory (PI) systems. We now know (and forecast may have been correct, and the supply and delivery
report later in this study), that PI systems are often functions executed appropriately. However, due to execution
inaccurate, and that the presence of “phantom inventory” is a in the store, for some reason the product didn’t make it the
major contributor to the 47 percent store ordering root cause. final 50 meters so it could land in the shopper’s cart.
In our 2002 study, we identified that on average 25

percent of OOS were shelf OOS, when the product was
actually in the store, but simply not on the shelf at the time
the shopper wanted to purchase the item.

Chapter 1
shown in Figure 4. At minimum, the cost to manufacturers

RFID and Shelf OOS on average is 35 percent of intended sales, and the cost to
The advent of the use of RFID in the industry has
retailers is 40 percent of intended sales. And from what was
brought increased attention to shelf OOS. Most RFID
presented earlier in the report, the total cost is much higher
technology is limited to tags on pallets and cases, and—due
than the lost sales.
to technological and financial reasons—has not descended
to the individual item level. Thus RFID is limited in the
Figure 4: Consumer Responses to OOS Events
ability of the technology to address shelf OOS. The major
Do not purchase item 9%
application for reducing shelf OOS is to understand what
Substitute
inventory exists in the backroom and to locate it, where
different brand 26%
hidden items are a continuing problem for store stockers.
RFID applications are focused on knowing when the

cases come into the store’s backroom, and when they move But item at
from the backroom to the store floor. It can also identify another store
when a case is returned from the floor to the backroom. As 31%
such this has implications for store stocking practices. For
example, this suggests that a case that cannot be completely Substitute
stocked on the shelf becomes a problem because the RFID same brand 19%
does not recognize a partial case in the backroom.
Delay Purchase 15%
Alternatively, RFID applications can help dramatically Credit: Gruen, Corsten, and Bharadwaj 2002
when sorting cases coming off a delivery truck. One might
consider a sort of “triage” (borrowing from emergency Worldwide Consumer Responses to OOS Events
medical care where the worst-case patients get the first Average across 8 categories
attention). Thus items that are known OOS get identified We also know how responses to OOS vary based on the
quickly for immediate stocking, while items that are still type of product. A group of Dutch and Belgian researchers
available to the shopper but have room on the shelf for a have provided a useful theoretical model to examine the
full case get secondary attention. Cases that are back-stock shopper’s opportunity cost, transaction cost, and substitution
remain in the backroom, rather than being taken to the cost (Campo and Gijsbrecht 2002). When the opportunity
sales floor and returned. cost of not being able to immediately consume the product
is high (for example when one runs out of diapers), the
Backroom Inventory Correlates consumer will either substitute or find the item at another
Positively with OOS Levels store. Alternatively, a low opportunity cost will lead to either
While dominant logical thought would consider larger purchase delay or cancellation. When the substitution cost
levels of backstock to reduce OOS levels, our research (as of using a less preferred brand is high (for example in the
published in the 2002 study and as updated in Chapter 3, case of feminine hygiene and laundry), the consumer will
Section IV in this report) has found the opposite to be true. take any action except to substitute another brand. When
The reason this occurs is that the back room of the retail the transactions cost is high (the time and effort to purchase
store should only store items where there is known demand later or elsewhere), the consumer will either substitute
beyond what can be kept on the shelf until the next delivery or cancel purchase. Each individual cost component is
from the DC will be made (for example promoted items). limited in its ability to explain the consumer response,
Any inventory in addition to this that is kept in the back however, different reactions can be robustly explained by the
room effectively creates a second storage point, which is interaction of the three components.
costly and inefficient by definition.
Another group of Dutch researchers examined a
1-4. We Know How Consumers Respond to OOS comprehensive set of antecedents to shopper OOS reactions
Conditions (brand related, product related, store related, situation related
and consumer related). They found that the brand and
Reactions to OOS items
product related antecedents had the greatest influence on
In general, our knowledge of how consumers respond
shopper reactions. From the results, they recommend OOS
when facing an OOS item is well-known. Our 2002 study,
reduction strategies based on the equity of the brand and the
based on 72,000 shoppers, published overall worldwide
way the product is used. While the overall conclusions are
benchmarks that have been generally accepted, and these are
10 A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry
Understanding the Total Cost of OOS
Figure 5: Shopper Cost Components and OOS Behavior
When the And the And the Then the Customer will…
Opportunity Substitution Transaction
Cost is… Cost is… Cost is…
High High Low Buy item at another store
Low High Low Delay purchase
High High High Substitute—same brand
High Low High Substitute—another brand
Low High High Not purchase item
Credit: Gruen, Corsten, and Bharadwaj 2002
generalized to entire categories and require refinement to be

diagnostic, the overall approach is enlightening and provides
good general guidance (Sloot, Verhoef, and Franses 2005).
There are still some critical things we don’t know about

consumers. Most importantly, we don’t have empirical
evidence of how OOS affects store choice, either initially,
or over the long run. We do know that when a shopper
has to go to another store to get an item that was OOS at
the preferred store, that the other store has been given an
opportunity to win additional business besides the OOS
item. Over time, business will erode due to OOS items, but
we do not know at what rate this occurs.
Conclusion
Our knowledge of consumers gives management plenty
of reason to consider reduction of category assortments
and reallocation of shelf space to faster moving items. Such
efforts result in fewer OOS as well as lower labor costs, as
faster-moving items do not need to be stocked as frequently.
The only surprise at this point is why this conclusion is not
being implemented. In Chapter 3, we present our research
on demand-based planograms, and this provides an initial
approach and rationale for retailers to provide optimal
assortment, as opposed to the widest assortment.
Gruen & Corsten 2008 11

Chapter 1
Understanding OOS
through Measurement 2
2-1. Understanding the Measurement and 1. the frequency that events occur over the measurement
Identification of OOS period;
2. the length or duration of each event within the
Defining OOS to Measure the Impact of OOS: measurement period, and the total cumulative duration
Events, Rates and Lost Sales during the measurement period, and the inverse of this
In order to address OOS, one must know what they which is the availability, which is the probability that a
look like, and thus be able to recognize and describe an shopper will be able to find the item on the shelf at any
OOS. How does one describe an OOS? How does one given moment;
calculate the extent and interpret the degree to which OOS 3. the intensity, or the loss to the organization due to
has affected their company? OOS, which has multiple related components including
the lost unit sales, and the lost monetary sales. Other
In the introduction, we provided a detailed definition derivative measures of intensity include the net sales
of an OOS, and we also provided a comprehensive list of loss due to OOS (after consideration of substitution),
the attributes that are typically measured to determine one the gross margin net sales loss, and the number of lost
of several OOS rates. In this section, we provide additional shopper baskets (a measure of number of customers that
detail on OOS attributes, with a particular emphasis on how are dissatisfied), accounting for multiple sales loss to a
each type of measurement of the attributes can be used to single customer; and
detect the root causes behind the OOS. 4. the breadth (number of stores with simultaneous
pattern), which indicates the likely location of the root
In our research we came across a variety of indicators cause (individual store, supply chain, manufacturer).
used to describe OOS, and the differences depended on
the ways that the OOS were measured. The definition of Best Practice Recommendation: These attributes must
what makes an OOS also affected the extent that has been be measured to determine the overall impact of OOS. This
reported in studies. Thus, for this study, we took particular approach to identifying and measuring multiple OOS
care to clarify the definition that would allow us to measure attributes should be adopted as a best practice for the
OOS in a way that would provide adequate understanding to industry.
move towards solutions.
The “OOS rate.” As we showed in the Introduction,
Events and attributes. To begin, most will agree about the notion of an “OOS rate,” which we determined in our
the meaning of an “OOS event,” referring to whether the 2002 study to be approximately 8 percent worldwide, is
product was or was not on the shelf in a salable condition, a function of the occurrence of the measured attributes
as we discussed in the introduction. This is the starting over the measurement period. It is worth reiterating that
point. However, simple identification of OOS events does the traditional “OOS rate” that has been reported in most
not provide adequate diagnostic understanding that will studies, has been measured through manual audits, and thus
lead us to finding the root cause and a solution. With the indicates the number of items that are OOS at a given point
identification of each attribute of the OOS event, we find in time. This measurement has little regard for the duration
that we can better identify the root cause and more efficiently of the OOS, or the degree to which the OOS impacts the
and effectively apply a solution to the OOS. To review retailer. As such, we recommend that whenever the term
the attributes listed in the introduction, we note that the “OOS rate” is used, that an adjective that indicates the
attributes are viewed in a specific measurement period (day, attribute being measured also accompany the term.
week, month, year), and from this we can derive various rates
that are typically expressed as a percentage. Here are the
relevant attributes we identified:

Chapter 2
The first is a manual auditing method where periodic

Best Practice Recommendation:
checks are made of the shelves or other merchandise displays
When referring to an “OOS rate,” the industry should in selected categories. This method formed the base of
adopt as a best practice the use of our clarifying most our findings in our 2002 study. The auditor counts the
terminology of the various rates explained in the number of “holes.” A “hole” is defined as an item that has
introduction. a shelf tag or otherwise should have inventory on the shelf
according to the planogram for that category, but there is no
Sales losses due to OOS. While understanding the inventory for that item on the shelf visible to the shopper. If
frequency and duration of OOS rates is a critical starting the item is placed in the wrong location or hidden behind
point, understanding of the sales losses provides the another item, that also counts as a hole. The total number
diagnosticity required for addressing solutions that lower of holes that are found in that given period are then divided
OOS in an efficient manner. Some OOS are more costly by the total number of items in the category to provide a
than others, and a focus on reducing the more costly OOS percentage of items that are OOS.
will lead to greater return on investment to lower OOS, than
simply lowering the overall rate without regard to which In the traditional and most commonly used approach,
OOS make a bigger difference. As this chapter demonstrates, the OOS rate is measured as a percentage of SKUs that
OOS of fast moving items encounter greater lost sales than are out of stock on the retail store shelf at a particular
do OOS of slow moving items. moment in time; i.e., the total number of items that the store
usually carries but are not available. Normally, the OOS
Lost sales occur when a product is not available at the rate is reported for each category individually and then the
time a shopper wants to buy it and as consequence shopper categories are averaged to create and report an overall OOS
behavior is altered so that the intended item is not purchased, rate.
purchased at a later time, purchased at another outlet, or
substituted with another item. Even when the item may be There are four primary advantages to this method of
somewhere in the store or otherwise be hidden from the measurement:
shopper, if the shopper is not able to purchase the intended • The method has been used for many years and there are
item, it is considered to be a lost sale due to OOS. When an reliable benchmarks available for comparison.
OOS event occurs, the item does not incur a lost sale unless • The method also produces results that are believable, i.e.,
it cannot be replenished to the shelf before affecting the next seeing is believing.
shopper that wants to buy the item. Alternatively, a single • In the aggregate, in the manual method, the OOS
OOS event can result in multiple lost sales. The approach rate will approximate the lost sales rate due to the
of examining the sales loss has major implications on how complementary errors where lost sales are over-
retailers prioritize and address OOS. estimated for slow movers and under-estimated for fast
movers, and therefore they balance each other out. Based
Best Practice Recommendation: on our research and comparison of this method with
others, the overall averages of OOS levels measured this
When seeking to address OOS, the OOS should be way tend to approximate the overall lost sales for the
prioritized based on obtaining some understanding of the retailer.
losses associated with each OOS event.
• It provides a cost-effective way to address on-shelf
availability in a known problem area of the store.
General Approaches to OOS Measurement
When examining OOS, there are three general The limitations to this approach to measurement
alternative approaches or methods of measurement: include the following:
A. manual audit method; • all holes are counted as equal, even though some may be
B. POS sales estimation; and for much faster selling items or may have much longer
C. perpetual inventory aggregation. durations;
• the audited OOS rate does not measure either the lost
Each method has its advantages and disadvantages, revenue or consumer impact of each OOS despite often
and—when used effectively—can direct managers to root substantial differences between them;
causes of OOS. • the selection of the categories is often unbalanced, being
influenced by a supplier who sponsors the audit;
A. Traditional Audit-Based Measurement of • the frequency and timing of the audits, generally
OOS—Looking for Holes determined by the overall budget provided for
Understanding OOS through Measurement
measurement and availability of staff, often misses peak shelf availability for all items (this is further addressed in
business hours and weekends and produces biased OOS Chapter 3).
rates – audits are often performed first thing in the
morning when the shelves are more likely to have been
Best Practice Recommendation:
restocked;
• the duration of the audit (typically a week, again When conducting a manual audit of OOS items, retailers
determined by budget) is often too short to measure the should seek to associate the findings with likely root
true average OOS rates because of normally large daily causes. This can be done by comparing the items that
and weekly variations in OOS rates; are found OOS with PI data, items being advertised or
• the audits are very labor intensive and can interfere with otherwise promoted, items kept in multiple locations in
routine store operations; the store, delivery frequency, shrink, ordering amount, etc.
• the accuracy is diminished by unavoidable counting and
other human errors due to inattention and fatigue of the B. Data-Driven Measurement: Estimates Based on
auditors – especially if auditing goes on for an extended; POS Data
• when a “hole” is filled with another product and the Instead of relying on physical audits, the second
shelf tag removed (for items the retailer intends to have approach is measured through the use of models that
in distribution), then the OOS cannot be identified, and estimate lost sales from OOS using store scanner and
thus OOS are underreported; inventory data. This approach to measuring OOS directly
• when the shelf is filled with the wrong item even when estimates the number of times a consumer actually intends
the tag is present, the OOS may not be identified if the to purchase the SKU and does not find it. The percentage
auditor relies only on visual inspection (but scanning the rate is calculated as the number of times the consumer does
item’s UPC could identify the error); not find the SKU divided into the sum of the times the
• the overall cost is prohibitive for sustained consumer does find the SKU plus the number of times the
measurement—the total cost is function of the consumer does not find it. Variations on this measurement
frequency and duration of the audit; and estimate the lost sales quantity and/or the lost sales revenue.
• this method is not feasibly scalable to a large number of This view provides the advantage of determining the extent
categories or a large number of stores. that OOS affects the retailer and the upstream supply chain
members. Moreover, the data used to make the estimates
This final limitation is crucial to any effort to lower lost also provides the duration and frequency of OOS. And, it
sales due to OOS. While the audit method provides reliable enables a much more detailed root cause analysis because the
averages of the OOS rate that—when consolidated at the start and stop time of day and day of week are identified for
store level tend to provide an overall measure with reasonable each OOS event.
accuracy—it assumes that each OOS item is equal in its
effect, and thus does not provide the diagnostics necessary This method of measurement using POS transaction-
to address OOS reduction efficiently. A slow moving item level data has been shown to identify true positive OOS
that has no inventory on the shelf, but may not have a at an accuracy of 85-90 percent when validated by manual
shopper come to purchase it for several days is counted as a audits. That is, 85-90 percent of the items identified as OOS
single OOS event, while a faster moving item that may have were in fact either OOS or had a shelf tag error. Because
multiple shoppers intending to purchase it in a single day is each OOS event is individually identified, OOS losses can be
also counted as a single OOS event. aggregated to brand, category, department, store, or any other
level of management interest, at the same degree of accuracy
Given all the limitations, the manual method can still (80-95 percent). One disadvantage is the higher (but
be valuable to the retailer by pointing towards major root currently unmeasured) false negative rate – i.e., the number
causes of the OOS. Often, addressing a root cause will of items actually OOS but not detected as being OOS.
lower the OOS event rate of all items that were affected by Because of the way these data-driven methods estimate
it, regardless of the sales losses. For example, OOS items OOS, false negatives tend to be items where the losses are
that are identified through a manual audit can be compared small.
with the perpetual inventory (PI) data for those items. If
the PI for the OOS shows a large number of the items to • The major limitations of this method include: the OOS
be on-hand, i.e., they are “phantom inventory,” then the PI rates are estimates based on historical sales patterns, and
inaccuracy would be the likely root cause for those items. when the past is a poor predictor of the present, accuracy
Improvements in PI accuracy will then increase the overall is a problem;

Chapter 2
• it does not work well for SKUs that sell slowly (thus are (taking into account a variety of variables including sales,
restocked without ever being detected as OOS ); prices, seasonality (time/day/week) as well as changes
• the method depends on accurate POS data; and variances to these.
• initial cost to set up; and 3. When an item’s purchase cycle (expected velocity) is
• the estimates are not immediately trusted by store interrupted beyond a calculated threshold, that item is
management, since they have been mathematically deemed “OOS.”
estimated instead of physically observed. 4. The item is reviewed to determine the likelihood for
“false positive” identification (estimated OOS when it is
In spite of these limitations, much work is going actually in stock).
forward using this method. It overcomes most of the 5. After the next actual sale of the item occurs, the total
limitations of the manual audit method, by focusing retail number of lost sales is calculated.
managers on the OOS items that are impacting their 6. The lost sales and relevant OOS rates (events, duration,
shoppers the most, and by taking away the labor intensity lost unit sales, lost monetary sales) are then aggregated
and human error of measurement. Furthermore, while into a managerial report.
there is an up-front cost for the method for the retailer, the
variable cost of analyzing additional categories or longer Figure 6 illustrates with two examples how the
periods of time is much less that physical audit methods. estimation process works. The first example shows three
lost sales (the first occurring when an expected purchase is
How does POS Estimation Technology Work? missed followed by two additional expected sales), and the
A variety of methods have been developed by second example shoes four lost sales (the first occurring when
commercial ventures, consultants, and academics that use an expected purchase is missed followed by three additional
mathematical algorithms to estimate lost sales based on expected sales).
historic patterns of sales of the item. Each method that we
reviewed has its own nuances, advantages, and disadvantages POS estimation holds the greatest benefit for retail
over other methods (in terms of its sensitivity, cost, accuracy, outlets where there are a large number of items where the
etc.) However, each mathematical algorithm tends to address product movement is faster than the typical replenishment
OOS by the following general steps: cycle. While OOS can be detected for slower moving items,
1. The algorithm examines each item’s historic sales these may be detected more quickly and efficiently through
velocity (normally requires a year). other methods.
2. It sets parameters for expected future sales velocity
Figure 6: Two examples of OOS Estimation Using POS Data
Credit: Data Ventures

OOS Duration receive historical reports, some retailers are working with
One attribute calculated by the POS measurement OOS identification vendors to obtain real-time notification
method is the duration for each OOS. Figure 7 shows of OOS as they are identified. POS data on the faster
the distribution of 1,084 OOS events on the 200 fastest moving items is continuously analyzed to determine when an
moving store items that were detected for a U.S. retailer OOS has occurred, and the retail manager can immediately
over a 14-day period. These duration measures include the address an identified OOS event. This also provides a way to
hours the store is closed. Figure 8 provides the cumulative validate the algorithm used to identify the OOS: how many
perspective of these events. This shows that 40 percent were times is an OOS that is identified by POS data actually
one day or less, and an additional 20 percent lasted two days. appearing as an OOS (i.e., a hole) when store personnel go
Particularly troubling is the fact that almost one-quarter to address it? The objective of real time notification of OOS
lasted three days or more. is to reduce the OOS duration
Examining duration distribution patterns helps point Figure 9 shows OOS duration in a study looking at 12
to likely causes. It also helps understand losses from OOS, months of data for 100 products in 24 stores. For this retailer,
since the loss from an OOS is a total of the lost sales for that the duration of these OOS items has a very different pattern
item over the period that it is OOS. Rather than waiting to than shown in Figures 7 and 8. The red bars (left bar in each
cluster) identify the percentage
Figure 7: Duration of 1084 OOS Events of OOS that lasted for that
particular period, while the blue
bars (right bar of each cluster)
identify the cumulative OOS.
For example, about 6 percent of
OOS last three days, while 10
percent last three days or less,
and about 25 percent last four-
to-six days, while 35 percent last
six days or less. There is a marked
impact from OOS episodes
with duration of one week or
longer. This data indicates the
opportunity available to this
retailer for these 100 items: if
store operations could simply
Credit: Standard Analytics indentify and correct all OOS
occurrences lasting seven days
or more, then this would reduce
Figure 8: Cumulative OOS Duration from Figure 7 overall OOS rate by more than
60 percent.
Differences in the duration

patterns between the two
retail formats suggest that the
root causes and thus the best
solutions will be different for
each format. As the overall
theme of this report suggests,
the key is to be able to get
good measurements of the
OOS attributes, link these
measurements to the likely root
causes, and then address the root
causes with focused resources.

Chapter 2
C. Measuring OOS Using Perpetual Figure 9 – Another View of OOS Duration

Inventory (PI) Data
The third method for measuring OOS is using PI data.
PI systems track sales, and when sales = 0, the item is OOS.
While retail formats with faster moving items will benefit
for POS estimation methods, lower volume store formats
with a preponderance of slower moving items, may benefit
by measuring OOS using PI data. The major limitation
with this method that is described in Chapter 3-3 is that
PI accuracy is generally below 50 percent. Thus a manual
process for checking PI accuracy and potential OOS items
that are not indicated by the PI system must be implemented
simultaneously. Also PI data usually applies only to store
OOS and not shelf OOS, so it cannot directly indicate shelf
availability.
Credit: T3Ci
2-2. Focus on the Items that Result in the Majority of
OOS: There Really Aren’t That Many Fast Movers Figure 10
New analyses of POS data provide us with a clearer
picture of product movement across time. The conclusion: a
relatively small number of items constitute the majority of
the store’s total sales.
Our analysis of product movement shows that in a typical

day, large volume U.S. grocery stores carrying more than
50,000 items (SKUs) will sell approximately 12,000 different
items. This is pictured in Figure 10. On the busiest shopping
day of the year, the store will sell about 16,000 unique items.
In a single week, a store will sell about 27,000 unique items.
Looking at the chart from another perspective, even on

its peak day, only 5,000 items (about 10 percent of the total Credit: Standard Analytics, 2006
SKUs carried by the store) accounted for over 70 percent
of the store’s sales. For a typical week, the classic 80-20 rule
applies: 20 percent of the store’s items (about 10,000), account Figure 11
for 80 percent of the store’s sales).
This same pattern is even more pronounced for stores

with lower overall sales volume (Figure 11). In this case, only
5,000 unique items are purchased in a single day, and on the
peak day, about 7,500 unique items sell. In a week, 15,000
unique items are purchased, and 7,500 items constitute 80
percent of the store’s sales.
This general pattern shown here has long been recognized

by stores, but this new data analysis clearly demonstrates the
extent to which a relatively small number of items influence
the overall store sales.
Credit: Standard Analytics, 2006
Previous research has shown that OOS rates demand groups account for only 3 percent (363) of the items,
correlate strongly with item sales velocity. Thus, the faster they account for 17 percent of the total lost sales ($50,094).
moving items (including promoted items) constitute a Additional detail of the analysis and charts that show the
disproportionate share of OOS volume sales loss. Therefore, relationship between sales velocity and lost sales are provided
for the highest impact, the focus for reducing OOS needs in Appendix 1.
to be on the SKUs that make up 70-80 percent of the store’s
sales. Why is this understanding of product movement
important to the study of out-of-stocks?
The relationship of product movement The relationship between product movement and
to lost sales lost sales can have implications on forecasting, inventory
Since a relatively small number of items account for (ordering and keeping backstock of fast moving items), and
most of the store’s sales, the next question involves the merchandising (allocation of shelf space to faster moving
linkage between the sales velocity and the number of lost items). Moreover, this has specific implications for suggested
sales. To answer this question, we analyzed the movement solutions to OOS once the fast moving items have been
of items in a single store. We selected all the packaged items identified. Strategies and solutions for reducing OOS levels
that had appeared in a minimum of 180 baskets in the year, for the fast-moving items should be different than the
and that had scanned in at least 12 different weeks of the strategies and solutions for slow moving items. Sales rates
year. This was 11,407 items of the nearly 70,000 unique items of the identified fast moving items need to be matched with
that had scanned at least once during the course of the year. ordering and delivery cycles. For example, this analysis would
We separated the items into three groups. The first group was indicate that a single weekly delivery should be adequate
the constant high demand items, packaged items appearing to keep the vast majority of SKUs available on the shelf
in at least 70 baskets per week in at least 12 weeks of the year (without the need for safety stock), while there is a small
analyzed. The second group was the temporary high demand number of items that need more frequent deliveries (or
items, packaged items scanning in at least 12 weeks of the significant safety stock).
year analyzed, and with a lift of at least four-times above
their average weekly velocities in a given week. This group is Another implication of this analysis is that a retailer
mostly promoted items. The third group was all other items can not be ignorant of which SKUs are the most important.
meeting the minimum inclusion criteria, but not moving fast Several retailers we studied understood this to some degree,
enough to be included in the high demand groups. and they have instituted “never out” lists, or some sort of
special identification on shelf tags. However, the solution is
Figure 12 shows that there is a clear linkage between to understand the SKUs that are selling and accounting for
product movement and lost sales. While the combined high most of the sales, and as a result for most of the lost sales.
One example might be a
Figure 12: Number of Fast-Moving Items and the Corresponding Lost Sales “SKUs never out” strategy
employing the peak demand
multiple identified in Chapter
3. Once these are addressed,
the majority of lost sales
from OOS items have also
been addressed. A secondary
strategy for addressing the
OOS of slower moving items
can then be implemented
after the primary strategy for
lowering OOS from the fast
Credit: Standard Analytics, 2007 movers have been in place.

Chapter 2
2-3. Viewing OOS Patterns

Identification of OOS and Linkage to Root Causes
Beyond Measurement to Finding Patterns Analysis of OOS using this weekly calendar as a
Our previous research showed that OOS tend to form template has shown that there are some general patterns that
patterns such as day of week (for example see our 2002 study, OOS tend to form. Each pattern suggests a likely root cause
p. 14). These patterns can be identified through traditional (or causes), and thus the identification of the OOS pattern
shelf audit methods of measuring OOS. The research we points to a likely fix (or fixes). Four of these patterns are
have conducted for this study using POS measurement featured below.
methodology has confirmed these same patterns day of week
as well as time of day, and provides pictures of these patterns Utilizing patterns to address OOS:
in distinct detail (see the appendix for examples of this The general theme of this report is that sound
measurement). measurement of OOS needs to point to root causes, where
resources can then be directly applied as a solution. As
Using these simple patterns, managers can fashion the examples in this section show, identifying patterns of
strategies to address these known OOS patterns. For OOS items provides an intuitive way to link solutions with
example, knowing that OOS will increase on Saturday OOS items. This approach is clearly a breakthrough in the
and Sunday, stocking labor can be reallocated to increase identification and solution to OOS.
frequency of stocking during these periods. Similarly,
additional back-stock can be maintained to cover lack of Pattern 1: Promotional Velocity
weekend deliveries to the store. Underestimated (Figure 13)
The first pattern shows annual sales by week of a fresh
Given recent progress in estimating OOS from POS packaged good that is promoted (price) about one time every
data, it is possible to take the measurement to additional month. Promotional periods are indicated by the thin (red)
level of detail, and find additional patterns in the data using horizontal lines, and periods where the item is OOS are
the attributes of OOS of frequency and duration. By plotting indicated by the thick (black) horizontal line. Over the past
the duration of each OOS event on a weekly calendar (as year, the item has only experienced OOS during promotional
shown in Figure 13), one can obtain a visual representation of periods, and then normally for one or two days in duration.
the OOS pattern. This approach provides a new and effective Through understanding the specific OOS events during the
way to address OOS by identifying behavioral patterns that promotional periods, the retailer adjusted order sizes, and
may OOS tend to form. after January, the OOS events ceased.
Figure 13: Promotional Velocity Underestimated

Chapter 2
Pattern 2: Inadequate Shelf Restocking Frequency (Figure 14)

This pattern shows a series of short OOS occurrences that can occur any day of the week. This pattern shows that the item is
in the store, but not always on the shelf. To address this OOS pattern, the store needs to add shelf space and check the stock
more frequently and restock the item more frequently.
Figure 14: Inadequate Shelf Restocking Frequency
Pattern 3: Inadequate Replenishment Schedule (Figure 15)

The pattern here shows that the period of OOS occurs nearly the same week after week. In this case it looks like there are
two deliveries a week, but four deliveries are needed. As an alternative, the store could increase the safety-stock of this item.
Figure 15: Inadequate Replenishment Schedule
Pattern 4: Inadequate Shelf Stock for Daily

Demand (Figure 16)
Here the item is OOS almost every day, indicating that inadequate space on the shelf has been allocated to this fast-mover.
The solution is to substantially increase the stock on the shelf as well as checking the item multiple times during the day.
Figure 16: Inadequate Shelf-Stock/Shelf Space for Daily Demand

Chapter 3
Our Hypothesis and the Related Studies

Lowering OOS Rates:
3
Overview 3-1. Ordering What You Need When You Need It: The
In our research, we focused on seven general areas that Case for Item Data Accuracy
are outlined below. For each of these we developed multiple In this section we examine the degree to which
hypotheses regarding their impact on levels of retail OOS. the actual items tracked in the inventory and ordering
Our previous research suggested that each would have an information database of the retailer matches the items that
effect, and in this study, we sought to find the degree each are actually for sale in the store. (Note that section 3-2
had an impact, and in what situations each would play a examines the degree to which levels of inventory of each
greater or lesser role. item in the stores does not match the level of inventory in
the store’s database.) In general, one should expect that the
In this chapter we present the findings of our level of OOS will be lower the more that the product item
proprietary studies, others’ studies and additional evidence, data in the retailer’s database accurately matches the data of
and it is organized in the following seven sections: manufacturers’ current SKUs.
3-1. Product Item Data Accuracy A. Which Supplier and Retail Practices Lead to
Evidence from GCI studies and Data Synchronization Item Data Inaccuracy?
Data inaccuracy in retailers’ inventory databases comes
3-2. Demand Forecasting Accuracy from a variety of causes including:
The self-fulfilling prophecy • Merging previously independent databases; inaccuracy
Accounting for last sales in forecasts occurs due to mergers and acquisitions or due to the
joining of previously separate data systems.
3-3. Ordering and Inventory Accuracy • Introducing new or discontinuing old products; new
The impact of accuracy of perpetual inventory item information is not correctly recorded in or deleted
Other studies’ evidence of the impact of perpetual from the database. Also, manufacturers regularly
inventory accuracy make minor changes in many of their items (e.g.,
Best practice for addressing PI accuracy down-counting, which is the manufacturer practice of
removing a small portion of the product in the package,
such as reducing a 100 tissue box to 95 tissues). Most
3-4. Replenishment of these changes are subtle with one very similar item
Frequency replacing an existing item. Replacing the old with the
In-store processes new UPC/GTIN code for what is essentially the same
product is sometimes forgotten.
3-5. Merchandising: Shelf-space Allocation • Managing seasonal or temporary products;
Demand-based allocation vs. Case-Pack allocation manufacturers often introduce temporary product
Theory changes, such as bonus packs with a new UPC/GTIN
code, and then revert back to the old UPC/GTIN. This
3-6. Merchandising: Planogram Compliance can lead to data errors in the database.
Theory-Hypothesis
Measuring POG compliance—best practice B. What is the Effect of Improving Data
Findings from study of four categories Synchronization?
Small differences can have a large effect. Even a single
3-7. Merchandising and Shelf Management digit that is incorrect in a product code or description can
Item Management cause a mismatch that will lead to not making an order a
Findings from study necessary product, or making an order for an item that does
24 A Comprehensive Guide To Retail Out-of-Stock Reduction In the Fast-Moving Consumer Goods Industry Gruen & Corsten 2007 24
Lowering OOS Rates: Our Hypothesis and the Related Studies
not exist. These errors are costly to track and fix because
UPC to GTIN
they incur substantial labor costs which would often justify
investment in a solution.
January, 2005 was the industry “sunrise” date for
Third-party vendors have evolved to facilitate manufacturers and retailers to be “GTIN Compliant”
collaboration between retailers and manufacturers. They meaning that products should carry GTINs and store
function as a clearinghouse for the data, matching retail scanner systems were required to be able to read 14-digit
GTINs (Global Trade Identification Number). Previously
databases with product lists provided by the manufacturers.
they were only required to be able to read 12-digit UPC
This collaborative effort between supply chain partners has
codes. The purpose was to create a unifying bar code
been facilitated by the industry group, the Global Commerce
identification number readable in all areas of the world.
Initiative (GCI).
Many different standards exist in different countries of
the world, and the GTIN was designed to be technically
The effects of data alignment on lowering OOS can
inclusive of all the standards. Thus, where the USA had a
be substantial as the following two pilot studies reported by 12-digit UPC and Europe had a 13-digit EAN (European
Capgemini/GCI 2005 show: Article Number), both could be included in a GTIN. The
• Johnson & Johnson and Wal-Mart have eliminated all newer standard allows manufacturers to sell items around
OOS of J&J products that were caused by data integrity. the world without maintaining separate inventory with
This represents a 2.5 percent reduction in OOS. each of the regional identifiers, thus simplifying trading
• In Latin America (Mexico, Guatemala, and Columbia), dramatically.
The Procter & Gamble Company and several retail
customers reduced purchase order errors from 3.6
percent to 0.8 percent, and this resulted in a decrease in
OOS items from 8 percent to 3 percent. RFID versus EPC
C. Using Parent-Child Relationships to Eliminate RFID is the broad term for the technology where a
Data Synch Errors reader identifies (reads) tags as they pass in close proximity
One method of eliminating data synch errors is to the reader. The most public display of the technology
to establish parent-child relationships. SKU’s that are is at the exit of stores where they are used for theft
temporarily replaced by a package with a different UPC may deterrence and if a powered tag passes through the portal
an alarm goes off. Tags can be passive (i.e., the tag has no
be tied together using the retailer’s item code. If a retailer
power and only responds when the radio signal hits it) or
has an item code that is six digits long, the first five might be
active (i.e., the tag can send a signal on its own).
the base number with the last digit a zero for the base SKU
and different numbers to indicate a temporary pack. So, for
EPC is the set of standards established by GSI and its
example, if a Crest Toothpaste has an item code of 21342-0,
subsidiary EPCglobal for multi-industry use. The format
that code would be the base item. The same Crest Toothpaste
of the tags and the fields of information they carry are all
as a bonus pack of 25 percent free product might then have standardized under EPCglobal governance so that all tags
an item code of 21342-1 and a second bonus pack with a will contain the same data in the same format. The parallel
trial size Scope attached might be 21342-2. would be where UPC is a standards-based format of a bar
code used for the FMCG business (and others). There are
Using this system all the sales for the SKU can be other formats of RFID tags, just as there are other formats
captured and aggregated into a single SKU code number. of bar codes in other industries.
This system would not work for down-counting where an
item permanently changes from a 100-count box to a 95- An EPC tag is a unique identifier for the individual
count box. item the tag resides on. Thus, each package of Pampers
can be uniquely identified as opposed to a UPC code
D. Conclusion where all packages of the same SKU have the same UPC,
Industry initiatives that address data accuracy and the packages appear all the same to the UPC reader.
through data synchronization show that the industry can
address OOS through a coordinated collaborative effort.
Manufacturers and retailers need to participate in this
initiative to achieve the benefits.
Chapter 3
3-2. Ordering What You Need When You Need It: The stores in the chain) of a major U.S. retailer. Of the total audit
Case for Inventory Accuracy of more than 20,000 items, PI was accurate only 45.4 percent
In this section we examine the degree to which levels of of the time, while 18.8 percent of the time it was +-1 unit, 10
inventory of each item in the store match (or do not match) percent +-2 units, and 25.8 percent it was off by 3 or more
the level of inventory in the store’s perpetual inventory units. These findings are in line with published research
from a group of academics from University of Chicago and
(PI) system. (Note that section 3-1 examined the degree
the Harvard Business School. They examined nearly 370.000
to which the actual items tracked in the inventory and
inventory records of a large retail chain and found that
ordering information database of the retailer matches the
only 35 percent of all inventory records equalled on-hand
items that are actually for sale in the store.) In general, one
inventory (Raman, de Horatius and Ton 2001).
should expect that the level of OOS will be lower when the
actual level of inventory of each item in the store accurately
matches the level of inventory of each item in the PI system. Figure 17: Audited PI Accuracy of Major US Retail Chain
A. How accurate is inventory information?

Inventory record inaccuracy is a substantial problem for
retailers using automated inventory management systems.
Overall we found in our research that inventory records
match actual inventory on-hand at a surprisingly low rate. In
spite of large investments by retailers in perpetual inventory
(PI) systems that link to POS systems, physical audits
consistently show that PI data are typically accurate for less
than half of the items in the store.
Accuracy can work two ways, with the actual on-hand

inventory exceeding the recorded inventory level, or the
actual on-hand inventory lower than the recorded inventory
level. The physical audits show that about half of the time the
PI shows more inventory to be in the store than is actually
on-hand (referred to as “phantom” inventory), and that about
half of the time the PI shows less inventory to be in the store
than is actually on-hand (referred to as “hidden” inventory).
Thus inventory accuracy is typically recorded as +- 1, +-2, B. What is the effect of PI Accuracy on OOS?
+-3, and so on, reflecting the presence of both phantom and
hidden inventory. There are several sources for inventory The Effect of Phantom Inventory
inaccuracy. Some of these include: As we expected, the effect of phantom inventory on
• Improper scanning of items at checkout. Improper OOS was found to be substantial. In this study, when the PI
scanning of a single item results in a double error in PI, matched the actual inventory, OOS levels were 4.1 percent.
with one item too low and another item too high. However, for the remaining items where the PI did not
• Unknown (e.g., theft) or unrecorded (damaged match the actual on-hand inventory, OOS levels were 8.9
merchandise that wasn’t recorded) shrinkage percent. Thus OOS levels were about 2.2 times greater when
• Store stocker ability to make changes to “on-hands” in PI did not match the on-hand inventory.
the PI system
• Incorrect audits and adjustments It is important to note that the measurement of OOS
• Product lost in the backroom in this example is the physical absence of inventory on the
• Mislabeling of cases at the DC shelf, and not necessarily reflective of the lost sales. For the
• Improper handling of master data for promoted and retailer in this study, the vast majority of items are slow
alternate products movers, where the average non-promoted sales rate is 0.3
units per week. Thus including a factor for promoted items
The example shown in Figure 17 shows the results of a and considering typical ordering and delivery cycles, the
physical audit of 166 items (sampled from a total of 25,000 estimated lost sales at a 4.1 percent OOS level is more likely
items per store) in 121 stores (sampled from a total of 3,000 to be about 2.1-2.2 percent, not accounting for substitution.
Does this mean that lower volume outlets with slower C. Conclusion: the Effect of Inventory Accuracy on
moving items have less at stake than higher volume outlets the Level of OOS is Substantial
with a higher number of faster moving items? It depends Our research shows a very strong effect of PI accuracy
on the degree to which the PI does not match the actual on OOS levels (this is consistent with other studies
on-hands, the capacity of units on the shelf, the reorder point examining this effect, e.g., ECR-Europe 2003). The data we
set for the item, order and delivery cycle, and the rate of sales. received from multiple retailers show that overall PI accuracy
is surprisingly low, and thus there is a very large opportunity
Consider, as an example, a slow moving item that sells to lower OOS by improving the degree to which PI aligns
once every two weeks (weekly sales rate = 0.5 units), the with on-hand inventory. The cost of PI inaccuracy is huge,
order and delivery cycle is once a week, the normal shelf whether it takes the form of phantom inventory or hidden
capacity is two units, and the reorder point is one unit. The inventory. These costs should justify substantial investments
item can be reordered in single units, Such items can be in improved PI systems and enhanced retail store processes
reordered in single units with specialty retailers. If the on- that keep PI records accurate.
hand inventory = 1 unit, and the PI shows 3 units available,
when the one available unit sells, it will be OOS and the Lower volume retail formats with a preponderance of
PI will show 2 units available. Thus the product will not be slower moving items will particularly benefit from improved
reordered by the ordering system, and could be out of stock PI accuracy, as OOS events of slower moving items can go
for several weeks. This situation will continue indefinitely unnoticed longer than OOS events of faster moving items.
until a physical check of the shelf shows the item to be OOS. Additionally, slower moving items cannot benefit from the
Thus correction and reordering of the item will depend on signals from POS generated OOS recognition that focuses
a physical process, triggered as a regular shelf check, or by a on faster moving items.
customer who asks for the item.
D. Best Practice Recommendation to Increase
Alternatively, consider an example, a faster moving item Inventory Accuracy
that sells 1 unit a day (six -- seven units per week), has an Improvements PI accuracy do not necessarily require
order and delivery cycle two times a week, shelf capacity substantial monetary investment. Moreover, as in the
of 18 units, the reorder point is 6 units, and is ordered in situation we describe below, by better understanding PI
a case of 12 units. (Note: this it typical of a 1.5 casepack accuracy, this retailer actually lowered their labor costs
shelf inventory system used by many grocery stores.) If the associated with checking on OOS items while increasing
on-hand inventory is six units and the PI shows 10 units their PI accuracy. The steps used to increase PI accuracy
available, when four units are sold, the PI shows six units included:
which triggers an order for additional case (12 units). The 1. Establish current level of PI accuracy (for most firms
actual on-hand inventory is still two units, which would this will be less than 50 percent).
mean that with the bi-weekly delivery, that at most the item 2. Determine causes of PI discrepancies with on-hand
will be OOS one day. In this situation, an order would not inventory. For example, this retailer found the single
be triggered only when the PI inaccuracy is greater than the biggest driver of PI discrepancies was for products
minimum reorder point, in this case six units. Thus for this in multiple locations (e.g., end-caps, in multiple
item, PI inaccuracy of one or two items is of less concern planograms, at cashier, etc.)
than for slower moving items. 3. Examine the store processes used to count inventory,
and determine how effectively they address the leading
The Effect of Hidden Inventory cause(s) of PI discrepancy. For example, this retailer used
The effect of hidden inventory on OOS is not as manual cycle counts where more than 500 items were
damaging as phantom inventory and is indirect. In this case, checked weekly, but for 50 percent of these items, there
the PI system will trigger orders for additional product when was no change to the information that was already in
there is adequate supply in the store. Thus it is unlikely that the system. Moreover, the scanner “gun” did not identify
the store will run out of the item. However, the presence of multi-location items, which had been identified as the
unneeded inventory in the store not only includes the cost of major cause of PI discrepancy.
the inventory, more importantly it requires extra labor costs 4. Implement a new process to check most likely PI
to bring product to the shelf (which will often not have the discrepancies. The retailer implemented a new process
physical space for the additional inventory), and then return for store audits that included both what to check and
it to the back room (plus related costs of storing and tracking what not to check. The items to be checked were those
inventory and additional shrink). Regardless, given that that showed in the PI system that inventory was 0 or
hidden inventory carries substantial costs, it increases the negative, items where the PI was unusually high, any
justification for additional investment in PI accuracy.

Chapter 3
empty holes on the shelf seen by the store associate the inaccuracy is random in nature. They use a (Bayesian)
when walking the store, known problem items (such as probability distribution to estimate the “true” level of
high shrink items), items at the end of a promotional inventory on the shelves (DeHoratius et al. 2007). They
period, and the “red dot” never-out items. Items that created replenishment policies based on the statistically
should not be counted included new planograms, new updated inventory records to avoid the problem of “freezing,”
items, and items where there are one or more physically where a physical inventory position persists at zero while
on hand. the corresponding record is positive. In addition, they show
The results of the retailer’s Figure 18: PI Accuracy of 100 items across 24 Stores
approach to increasing PI
accuracy were impressive.
In a five-month test in 20
stores, overall PI accuracy
increased from 45 percent to
53 percent, and cumulative
accuracy within 1 item
(where PI is accurate or +- 1)
improved from 64 percent to
74 percent. At the same time,
by implementing the counting
process that directly addressed
the items with the greatest
likelihood of PI discrepancy,
they lowered the number
of total cycle counts by 20
percent.
E. Other approaches
to addressing PI
accuracy
We strongly recommend
correcting the inventory Credit: T3Ci
records through process
improvement or physical
auditing. However, similar to
the way that mathematical Figure 19: Root Causes for OOS across 100 stores, 500 products for 6 months
algorithms are being
developed to identify OOS
from historic POS data, new
methods and techniques are
being developed to enhance
PI accuracy using POS
and inventory data. The
researchers from University
of Chicago mentioned
above, for instance, recently
developed an intelligent
inventory management model
that accounts for inventory
inaccuracy, based on their
findings that such a large
percentage of the inventory
Credit: T3Ci
records are wrong, but that
that their replenishment policies recoup much of the cost

of inventory record inaccuracy, and that their audit policies
USING RFID technology to enhance PI accuracy
significantly outperform the popular “zero-balance walk”
audit policy. RFID technology can be used to support PI accuracy
by addressing a number of root causes. Case level tagging
Commercial providers have developed a similar can reduce or eliminate mislabeling and mispicking cases
approach. In Figure 18 shown below, the retailer’s PI at the retailer’s DC. Corresponding automated checking
accuracy was about 32 percent, where the algorithm’s when the cases (and tagged inner packs) are unloaded at
estimation of PI was accurate in 54 percent of the cases. the store can further improve the accuracy of flow data
This approach to improving PI accuracy will be helpful into the store. An important goal for RFID technology
particularly to larger format retailers with many fast-moving in the retail supply chain is to get a more accurate picture
and frequently replenished items. of backroom inventory versus sales floor (shelf ) inventory.
This approach first computes a corrected store inventory Product movement from the backroom to the shelf is
before estimating OOS levels. This permits subsequent monitored by RFID readers. However, at the tagged case
algorithmic processing that distinguishes Store OOS from and inner pack level interpreting this movement data is
Shelf OOS. An internal model of the retailer’s replenishment complex because partially empty cases may be returned to
algorithms permits further breakout of root causes. Figure the backroom and this cannot be sensed by RFID alone.
19 shows the results of this approach, and this reiterates One approach is to instrument the compactor used for
our general theme that measurement needs to point to root empty boxes in the backroom – a read of a case id at the
causes that can then direct resources to solving the OOS. compactor indicates that all its contents must have reached
the sales floor.
3-3. Demand Forecasting Accuracy
Our previous research established that 47 percent of The use of hand-held readers and multi-antenna
OOS events were caused by store forecasting. As this is the readers that are able to scan the entire backroom can bring
biggest chunk of root causes, we invested a lot of thought a more accurate and close to real time view of backroom
into better understanding this problem. Overall, the goal inventory. Continuous or periodic electronic inventory
here is to understand the degree that greater the forecasting counts are likely to be substantially more accurate than
accuracy drives lower levels of OOS. transactional approaches where even modest errors in
RFID scanning of additions and removals will quickly
accumulate.
A. OOS Create a Self-Fulfilling Prophecy in
Demand Forecasting Item-level tagging, which is starting to be used on
For most products, predicting sales using sales history high value items and time sensitive products such as
is difficult. Estimating demand accurately is even more video games and DVDs, brings clear inventory accuracy
complicated as the only reasonably accurate data that a improvements. With the appropriate instrumentation
retailer has at its disposition is point of sales store data. there can be electronic reading, not only of the precise
However, we know that previous sales do not equal the actual inventory on a “smart” shelf, but also verification that the
demand, if lost sales have occurred. In this case demand product is in the right place for a customer to find it. Item
equals sales plus lost sales. level tagging would also improve POS checkout accuracy
– however this is unlikely to be widely deployed until it is
Unfortunately, in grocery, lost sales are unobserved economic to tag most items in a store. In the apparel and
because most customers, who do not find the product other specialized retail environments item level tagging
that they intended to buy, make a decision to not buy, may already be economic because of the higher unit prices
buy elsewhere, or substitute, without registering the non- and the substantial labor savings in automated inventory
purchase of the intended item with the store. Even when counting.
asked at the store exit whether they found what they were Credit: T3Ci
looking for, many customers will not be able to recall
whether they found everything that they were looking for
unless they had it written on a shopping list because the Therefore, the inherent problem of estimating demand
impulse to buy a product may have arisen in the shop but with unobserved sales is that all models are imperfect
then immediately been suppressed by its unavailability. On because as true demand is unknown it is difficult to validate
the contrary, for instance in internet retailing, lost sales are how close to demand the proposed model has come.
observed as the system can monitor whether a customer Nevertheless, researchers have attempted to estimate demand
clicked on a product and then did not buy it because it was with unobserved sales. All models conclude that lost sales
not available.
Chapter 3
can be substantial and that it is strongly influenced by Interestingly, we encountered all types of retailers, from
average demand and demand uncertainty. We are hopeful those who always overruled the proposals to those who never
that the next generation of models will have solved this did. While all retailers had arguments for their specific policy,
dilemma by using longitudinal data which is currently hard recent research by the Technical University of Eindhoven’s
to find because of the cost associated with creating such a Retail Operations Unit found that store managers overruled
dataset. order proposals to balance the workload at the store by either
shifting the order to another day or by consolidating the
B. New Forecasting Models Must Attempt to orders into larger case pack sizes. They conclude that the
Account for Lost Sales store manager’s order adjustments provide valuable input into
From a more practical point of view, many software better calibrating the CAO system to store specific demand
companies have developed forecasting models. However, patters and suggest that case pack sizes matter.
most of them are fraught with conceptual problems. Some,
for instance, do not include estimations of lost sales and We also studied the ordering for advertising promoted
simply forecast future demand on the basis of historical sales. products of a major drug chain in the U.S. In this case, a
Others develop forecasts at aggregated levels, for instance, recommended order for the promoted items was provided to
on the distribution centre level or for product groups, which each manager who could then make ordering adjustments,
invariably leads to inaccurate forecasts for specific products at either up or down. The recommended order program
specific stores. incorporated a range of expected demand variability into a
statistical (Poisson) model, where along with other factors, it
Progress in mathematics, statistics and computing determined the level of safety stock for each store.
power now allow daily, real-time forecasting for on the
item and store level. This has enabled a surge in computer The study compared the managers who ordered more
automated ordering systems (CAO). A leading drugstore than the recommendation, less than the recommendation,
in Europe that has implemented sophisticated, causal CAO and did not change the recommendation. The study found
systems claims that it has been able to reduce out of stocks that overall managers tended to over-order relative to the
and store inventory substantially. Our own research with a actual realized demand of the promotion. Thus, managers
major European retailer supports these findings. In a field spent time seeking to improve on the CAO recommendation
experiment in 10 stores manually measuring the OOS rates while simultaneously increasing investment in unproductive
for 100 SKU over a period of 2 weeks we found the overall inventory. The conclusion from this study is to rely on a good
order related OOS rate to be 4.7 percent. However, upon system rather than people. Collectively, several thousand
closer examination we found that for those categories where managers make poorer decisions than the system.
products were manually ordered the OOS was 11.7 percent.
In contrast the OOS rate in the categories that used an D. Reaching the Ideal State: Sales Forecasting and
automatic ordering and replenishment system with simple Demand Forecasting Become the Same
heuristics was only 3.1 percent, a reduction of OOS of more One technology that should help improve demand
than 70 percent. These findings resonate with the experience forecasting (thus lead to ordering accuracy relative to
of the European drugstore that reports that the introduction demand) is the use of POS data. This assists ordering
of a new automatic ordering and replenishment system has managers by showing the speed at which items move, and
lowered their OOS rate by 70-80 percent. provides needed data to show changes in demand for an item
relative to changes in price or other promotional activities.
C. Research on the Effect of Store Manager
Discretion in Ordering and Forecasting. However, the presence of OOS items clouds this
Should Managers be Allowed to Adjust CAO picture. POS data only measures what customers actually
purchase, but not what they intended to purchase if the
Recommendations?
item is in stock. OOS items distort the true demand in the
There has been a long debate whether store managers
following ways:
with their knowledge of local customer behaviour can
• When an item is OOS, it cannot be purchased, and thus
do a better job in predicting demand than headquarter.
the sales are truncated in the POS data, showing less
Research in CAO seems to solve the dilemma how much
than the true demand. This leads to subsequent under-
discretion the store managers should be given in ordering
forecasting of the item.
and forecasting. Modern CAO system provides reasonably
• For the 45 percent of the time that shoppers substitute
accurate proposals to the store managers who can then
another item for an OOS item, this switching behavior
overrule the recommended order if they wish.
inflates the sales of the items that are in stock (beyond
their normal demand), while it attenuates the sales of

the OOS item (below its normal demand). In either
Silver Cleaning-Polish and Thanksgiving
case, the POS data inaccurately reflects the true demand Holiday Shopping: A Real Life Example
for both of these items.
• For the 15 percent of the time that shoppers delay their On the fourth Thursday of each November in the
purchase until later, the intervals of purchase of the item USA, nearly everyone universally celebrates Thanksgiving.
shown in the POS data become biased. In this case, the Traditionally, families and friends gather at homes for the
presence of OOS items will create a situation where the annual formal dinner. Grocery stores stock up on plenty
actual sales rate is more “lumpy” than the true demand. of turkeys, sweet potatoes, stuffing mix, pumpkin pie
ingredients, cranberry sauce and the host of other foods
What then are the implications when lost sales are that compose the traditional dinner. Given that millions
not considered when making forecasts? of homes serve a similar fare on a single day, surprisingly
• Sales forecasts are based on wrong parameters and few of the staple items are OOS in the stores, even on
are not set correctly (particularly for fast movers and Wednesday right before Thanksgiving Day.
promoted SKU), and this situation perpetuates lost sales.
• Category Management measures such as promotion and Carol went shopping on Tuesday before Thanksgiving
new product planning are based on wrong assumptions. to get the final items she needed to serve her 20 guests.
A product may be delisted because of poor sales when it Since the silver serving dishes and utensils would grace the
is simply often out of stock. Conversely, a product may table, she had silver cleaning polish on her list. The store
be promoted because it exhibits sales which are inflated carried one SKU, as a convenience item, and on this day it
by substitution effects from stocked out SKUs. was out-of-stock, since Thanksgiving also functions as an
• Inventory systems work with wrong mean and standard occasion for annual silver polishing. The silver needed to be
deviation and therefore ordering parameters (safety polished; so on her way home, Carol stopped at the drug
stock, lead time, supply frequency) are set incorrectly, store, and found silver cleaning polish in stock.
particularly for fast movers and promoted SKUs).
• Shelf facings are not set correctly. Next year, Carol will probably need to buy silver
• Case pack sizes are not set correctly. polish again, and while she is at the grocery store she’ll
• This in turn can lead to further lost sales, loss of brand probably check to see if it is in stock, and if not, she’ll
equity, inefficient supply chain and store routines and stop by the drug store again and get it there. And across
wrong allocation of store staff. the USA, there are thousands of others, just like Carol,
who will do the same thing. What the drug store hopes
How do we address this problem? is that the grocery store doesn’t ever understand its true
Currently, no techniques exist to reliably and validly demand, because the grocery store has no idea how many
estimate demand under unobserved lost sales. However, new units of silver cleaning polish it might sell. Given the lack
measurement techniques that estimate the lost sales during of demand data, the grocery store will continue to under-
a period or OOS based on historic POS data do provide an order and run out. Meanwhile, the drug store will order
important first step in approaching this problem. One can enough to satisfy its regular demand plus the spillover
assume that the lost sales estimated by historic POS data from sales lost by the grocery store. If the grocery store
is less than the true demand for the item, since the data could know its true demand and order accordingly next
that was used to estimate the lost sales already included an year, the drug store would be stuck with a lot of extra
unknown number of lost sales. So the question becomes: how inventory. But by the time the grocery store figures this
much should be added back into the forecast? out, Carol will stop looking for it at the grocery store and
Our recommendation for the amount is provided in the just go to the drug store to buy it.
solutions chapter, but the general approach is an iterative
process where increased order sizes by a factor determined by
the item’s sales velocity and volatility will reduce the amount 3-4. Having the Goods You Need When You Need Them
of error in subsequent time periods. Getting it completely As we reported in our previous research, both store
accurate cannot occur with current technology (after all, and shelf replenishment have been found to be among the
this is forecasting, not history), but we can remove a portion biggest causes of out of stocks. OOS associated with store
of the inaccuracy leading to more accurate forecasts that replenishment practices include late changes in promotion
contain less noise from lost sales. decision (after the order is placed), supplier policies that
encourage retailers to wait to place an order, improper cycle/
delivery times, as well as other upstream causes. In nearly 25

Chapter 3
percent of the cases, inadequate shelf replenishment was the sample and cannot be statistically projected to represent all
cause of out of stock—this is while the product is actually retail stores, it suggests that perhaps paradoxically, the larger
somewhere in the store. We found that the frequency with the backroom, the higher the level of out of stock. While this
which shelves need to be replenished is a function of many seems to run counter to traditional inventory theory, it can be
factors, including average sales and sales uncertainty, case explained by poor back room storage management. We have
pack size, shelf facings and capacity, lead time and desired found very often that overflow goods were almost randomly
service level. assigned backroom storage locations and no central system
tracked the movements of these goods. Typically, in larger
In this section we explore several replenishment-related stores, staff would spend a disproportionate amount of time
factors that drive levels of OOS. These include delivery searching for goods and often would not bother searching
frequency and timeliness, delivery fill rate and accuracy, or would only replenish items that were easy to see and
receiving accuracy, and backroom inventory levels. For the locate. Another clear sign of the lack of attention to the back
final factor, we recommend that best practice thinking room was that most store managers we asked could give us
considers a disciplined process where the retailer carries precise estimates of the sales area size but few could tell us
no back stock at all, and then makes allowances for certain the corresponding size of their backroom. The architecture
items and times (promotions and seasonality) where some drawings just gave one overall number for backroom
back stock may be required. Additional discussion of having including offices, bathrooms, changing rooms and storage
adequate stock on the shelf to meet demand between rooms, and the actual backroom is not treated as an asset (as
delivery cycles is found in section 3-5. is the on-floor selling space).
A. Replenishment Figure 20: The Relationship between Store Backroom Size and Store OOS Rate
Breakdowns: to the Store
and to the Shelf
A useful distinction can be
made between “delivery to the store”
in case of a sizeable back room and
“delivery to the shelf ” in case of
a small or no back room storage,
as well as between goods that are
delivered from a central distribution
centre (typical for most stores and
categories in Europe) or that are
delivered directly by the supplier
(direct store delivery or DSD).
In stores without backroom

storage, shelf replenishment
is tightly coupled with store
replenishment. When goods arrive,
they are unloaded from the truck Credit: Angerer 2006
and immediately carted to the shelf.
Often this action is executed by dedicated staff who are
available only for those few morning hours. During normal B. Centralized Delivery vs. Direct Store Delivery
sales hours, if a shelf is completely depleted, it often goes Regarding the distinction between centralized deliveries
unnoticed. And even if a depleted shelf is noticed, it cannot (DC) from retailer and direct store deliveries (DSD) from
be replenished as there is no backroom storage available to suppliers we found that the former was associated with lower
replenish it. out of stocks because store deliveries and subsequent shelf
replenishment were coordinated. On the other hand, direct
In stores with backroom storage, typically larger stores, deliveries by the suppliers were typically un-coordinated
shelf replenishment is de-coupled from store replenishment. with the shelf replenishment routines because the suppliers
In new research with a major European retailer we correlated relied on external merchandiser or rack jobbers to replenish
the OOS rate for 10 stores with the size of their backrooms. the shelves. Typically, the goods were put into backroom
While the regression line in Figure 20 is derived from a small storage upon delivery where they waited for the supplier’s
merchandisers to appear. In a large U.S. retail chain, Figure 21 Costs of Handling Products
DSD posed the biggest single problem to reducing OOS
not only because of the de-coupling of store and shelf
Inventory Store 7%
replenishment, but more that the retailers found it hard
Inventory Warehouse 5%
to impose discipline upon the supplier’s merchandisers to Transportation 22%
implement store policy.
C. Shelf Ready Packaging and Store Friendly

Delivery
It does not come as a surprise that many retailers,
particularly in Europe, have adapted their systems to
Handling Warehouse 28% Handling Store 38%
couple replenishment processes and to introduce shelf-
ready packaging as a new enabler of flow replenishment
from the distribution center to the shelf. Credit: TU/e Retail Operations Group
Recent research by the TU/e Retail Operations Group

across several retailers has found that the cost of handling
product in the store accounts for 38 percent of the supply Using RFID technology to improve
chain cost in retailing (Figure 21). Because of this high receiving accuracy
cost of store handling, retailers are increasingly interested
in ways that ease shelf replenishment such as store friendly
RFID technology can be used to support receiving
delivery and shelf ready packaging. Store friendly delivery
accuracy and backroom management. RFID can improve
describes methods to load pallets and products onto trucks in
the tracking of cases and pallets as they are received from
product groups reflecting the store layout so that they can be
the delivery truck. Cases can be quickly sorted into known
offloaded and transported to the shelves at once to minimize
back stock vs. those going directly onto the shelf. When
in-store transport and handling. Shelf ready packaging (SRP)
inventory is received accurately, inventory data accuracy
refers to products that come in a ready merchandised unit
(PI accuracy) will be increased.
which is easy to identify, easy to open, can easily be put onto
the shelf and disposed of, allowing an optimisation of shelf
replenishment and enhanced visibility. SRP covers all types
of shelf ready packaging, including promotional displays, 3-5. Making Room on the Shelf: The Case for Demand-
pallets. SRP can in many cases, impact the cost structure Based Planograms
of a product, since it may require industrial investment or In section 2-1, we showed how lost sales due to OOS
additional outer packaging cost. However, pilot projects correlate with product sales velocity. Because of this, many
by a group of European retailers including ASDA, Tesco, retailers have instituted a “red dot” program where the top
Sainsbury, and Carrefour together with many suppliers and identified “never-outs” are identified with a specially-marked
service providers have shown that SRP significantly reduces shelf tag to alert store personnel to watch for potential OOS
handling cost and increases in-store productivity, on shelf situations with those items. However, we view a “red-dot”
availability, and product recognition on shelf by the shoppers program as a workaround for a bad design. It works as a
(ECR Europe 2007) medication to soothe the problem, but is not a cure. The cure
is a redistribution of shelf space based on demand rather
D. Conclusion than case-packs. In this section we examine the degree to
Overall, we found a tight coupling of store and shelf which demand-based planograms (POG) will result in lower
replenishment to be a driver of on shelf availability. However, OOS than packout-based POGs.
this requires a more integrated design of shelf space and
delivery frequencies than what we have often found in A. Current Situation: 91 percent of the SKUs are
practice. Allocated Shelf Space Based on Case-Pack Size
In our examination of several categories across several
retail chains, we found that on average, 86 percent of the
items had enough inventory on the shelf to last more than
7 days (days of supply, or DOS). This means that only 14
percent of the items need to be stocked more than one-time
per week. Looking at the other end of the tail, several items
had in excess of two weeks supply on the shelf. If deliveries

Chapter 3
come weekly, the logical question to ask is, “why not give the Using these assumptions, and using a movement of
14 percent of the items more space, and take away shelf space one unit per week average for the slow mover items, then
from the slower-moving items?” The answer is that it is more in a given week, the fast movers would sell 140 units and
difficult to do than it seems, and this is due to two reasons. encounter 17 OOS events, while the slow movers would
First, most (91 percent) of the SKUs are allocated based on sell 14 units. Given these assumptions, the retailer would
case packout, and this takes up most of the space on the shelf. gain three weekly sales in the category. This appears to be
Since labor is expensive and tracking partial cases in the back somewhat equivocal, until considerations for the total cost
room is a nightmare, stocking a full case on the shelf makes of OOS are considered, as well as the reduced labor and
sense. Second, there is no additional space, since a slow tracking of 14 fewer SKUs. Alternatively, there are some
moving SKU cannot have less than one facing. customers who wanted to buy the 14 units of the slow
movers that are no longer available.
An empirical study at various retailers conducted by the
TU/e Retail Operations Group showed that retailers have Thus, before embarking on wholesale change, each of
what is termed “Net Shelf Space” (NSS). NSS is calculated the assumptions has to be understood. For example, if the
as the difference between the shelf space that is required to sales velocity ration between the fastest and slowest movers
carry out the current operations with respect to customer was increased, then the payoff would be greater. Alternatively,
service and costs and the shelf space that is allocated to the if the OOS rate on the items was decreased, then the payoff
item or items of interest. To examine the potential excess would be lower.
shelf space (ESS) an item might be allocated, the researchers
considered the merchandising guidelines of the retailer as C. Feedback-based Approach to Reallocating
well as the target shelf space advised based on a formula Shelf-Space
they derived for the maximum inventory level on hand. Given the above discussion, a very simple solution
As expected, case pack sizes, physical dimensions of the appears. When considering the 14 fastest and slowest moving
consumer units and the shelf depth were the major drivers of items, the results may be equivocal. However, if only a few
ESS. When the ESS of the items is summed, retailers were of the fastest moving items with the greatest OOS lost sales
found to have substantial NSS. The existence of NSS implies were identified, and additional space was found for these by
that there is space available to reallocate to items that do not eliminating a few of the slowest moving items, the results
have adequate shelf space. would generally be overwhelmingly favorable.
We also know from multiple tests in category Using OOS estimation using the POS estimation
management that shoppers respond favorably to minor method, this could be easily implemented. The top three
reductions in choice within a category, especially when there to five OOS items could be identified, and additional shelf
are reasonable substitutes. Thus, an option worth examining is space could be allocated to these by eliminating a few slow
finding additional space for fast moving items at the expense moving items. A month or two later, the OOS estimation
of removing the slowest moving items. The question is how could be made again, and the top OOS items would be
deep can a retailer make this adjustment until the negative examined. If it were any of the original three to five OOS
effects outweigh the sales gained from fewer OOS events? items, then additional space would be added to them, and if
not the next items would be considered. Given that there will
B. What Would be the Effect on Sales When 14 be diminishing returns, this would continue only until the
Fastest-Moving Items Receive More Space, and OOS reductions began to meet the sales level of the slowest
14 Slowest-Moving Items are Removed? moving remaining items.
Using the 86 percent DOS > seven days average, if the
category had 100 SKUs, then a reasonable place to begin to The approach here is to address the items that create
examine the effect of reallocating shelf space would be to the most OOS. These items can be easily identified using
find additional shelf space for the 14 items that did not have POS estimation, and implementation would involve minor
enough shelf capacity to last seven days by removing the 14 planogram redesign.
slowest moving items. Let’s make a few assumptions:
• The average product dimensions and the price margins D. Considering the Potential of Demand-Based
of the 14 fastest and slowest movers will be the same. Planograms
• The ratio of sales between the fast-movers to slow Most available computer software ignores peak demand
movers is, on average, 10:1. and demand variability. Planogram software by popular
• The OOS rate for the fast-moving items is 12 percent providers such as The Nielsen Company are based on mean
(1.5 times the worldwide average for all items). demand.
To show the potential impact of demand-based colleagues in the USA provided clear empirical evidence that
planograms, using a Data Ventures created algorithm, we assortment reductions—to a point—could not be perceived
examined three categories, looking at the weekly average by the shopper. According that research and subsequent query
movement for each item, as well as the regular peak demand (see Broniarczyk and Hoyer 2006) the focus needs to be
for each item. Building a demand-based planogram requires on providing an optimal assortment, rather than the largest
knowledge of the average movement of each item, as well as assortment.
the variability of the demand for each item. This variability
can be expressed as the multiple of peak demand versus the F. Conclusion
mean demand of the item. Those with high movement and There is a clear opportunity to consider moving towards
high multiples have the greatest likelihood of going OOS, demand based planograms. The increase for each category,
while those with slow movement and low multiples will have when examining recovered lost sales of the fast moving items
very low incidences of OOS. that receive additional shelf space, does not appear to be great.
The following chart, Figure 22, shows the ratio of peak However, when examining this in the aggregate, the
sales to mean sales, which indicates those with the highest revenue looks more attractive. For example, if the gross
demand variability. In each case, there is a clear “bend” in the margin for an item was only $0.25, and the net sales increase
curve where a few items appear to be at greatest risk. This would be only 10 items per week, the annual gross margin
chart shows that there can be a difference of up to 20 times, per category would be $130. However, the cost of OOS
or 2000 percent, between mean demand and peak demand. is much greater than the lost sales, and when the reduced
The demand-based planogram will provide additional labor for employees and customers is considered, the return
shelf space for these items. The peak is 5-6 times the mean is substantial. Moreover, because this method relies on POS
demand. For items that get above a multiple of 10, it is a estimation, the approach is scalable, so that the consideration
planned OOS. At the bottom end, there are so many things can be not for a single category for a single store, but for
that sell so regularly, that the space allocated to them could be several categories across hundreds of stores.
reduced.
3-6. Knowing Where Everything Should Be:
Figure 22: Ratio of Peak Demand to Mean Demand for Three Categories
The Case for Planogram (POG) Compliance
In this section we focus on merchandising

management, and we report on research that
examines the relationship between planogram
(POG) compliance and levels of OOS. Overall,
we found significant effects of planogram
compliance, however, in most cases we
examined, POG compliance was greater than
90 percent, thus the statistically significant
difference was not a large practical difference.
A. Findings from the Study

Credit: Data provided by Data Ventures In our new research we found on average
we found a relationship of 1/10 of 1 percent
change in OOS levels for every 1 percent change in POG
compliance. Surprisingly, we found POG compliance to
E. Reactions to assortment reductions be very high (greater than 90 percent) in the retailers we
If assortment is reduced, do shoppers see this as more of studied. For retailers and categories with high levels of
a negative issue than lost sales? Study after study suggests that POG compliance, there will not be much effect to address
a bigger assortment is not necessarily better. The early ECR this. However, for categories that have a low level of POG
studies in the USA showed consistently that when assortment compliance can have a substantial effect. For example, if a
was reduced by a small amount, that sales either remained category has 60 percent compliance, and it can be raised to 90
constant or increased. Moreover, customer satisfaction either percent, then there will be the opportunity to decrease OOS
remained the same or increased. In research published in by three percentage points.
the Journal of Marketing in 1998, Susan Broniarczyk and her

Chapter 3
B. General Issue of POG Compliance single grocery retailer) over a 14 week period.
While the previous studies have addressed getting the After the completion of the 14-week period, we obtained
product to the store, this study addresses merchandising measurements of OOS levels for that 14-week period for
processes in the store. When a planogram is created for a each category in each store. The OOS measurements were
retail category, space and position of goods has been allocated estimated by Standard Analytics using the POS data provided
based on known physical dimensions of the items and the by the retailer for the 14-week period and the previous 52-
number of items in a case pack. As such, the planogram weeks. We examined the correlation of the POG compliance
optimizes the shelf space utilization given consideration of and the OOS levels for each category in each store for the 14
stocking efficiency. week period. Two examples of the correlations are provided
in the charts that follow (Figures 23-24). The strongest
POG compliance first involves having the items in the correlation between planogram compliance and on-shelf
category that are supposed to be there termed
“in distribution”. However, some items that are Figure 23: POG Compliance and On-Shelf Availability in the Laundry Detergent Category
not specified on the POG (not “in distribution”)
find their way to the shelf, while other items
that are listed on the POG are not found on the
shelf. Secondly, POG compliance involves having
the items in the category in the correct place on
the shelf. Incorrect placement occurs when the
incorrect number of facings is provided for an
item, when the item is placed on a different shelf
rather than the one specified, or when the order
of the items on the shelf is different than the
order specified. Simple categories may carry 75-
100 items, while complex categories carry 300-
400 items. Thus the opportunity for compliance
deviations is huge.
The link between POG compliance and

OOS is both direct and indirect. An item that is
in distribution but does not have a physical place
on the shelf (i.e., a distribution void) cannot
be tracked for OOS manually. Similarly, an
item that is physically present on the shelf but Figure 24: POG Compliance and On-Shelf Availability in the Diaper Category
is not on the POG could go OOS and not be
recognized. Indirectly, POG compliance impacts
OOS due to the ability to easily physically
recognize low stock and OOS situations on the
shelf. Compliance also maximizes restocking
efficiency and accuracy.
C. Overview of the Study We

Conducted
To examine POG compliance, we identified
four categories (diapers, laundry, femcare, and
hair care) to study. We first had to establish
a “best-practice” methodology for measuring
compliance, as there was no commonly agreed
method to do so prior to our study (see the
following section on Best Practice Measurement
of POG Compliance for details). We used this
approach and measured POG compliance in
each of the four categories in 10 stores (of a
availability we found was for the laundry category (correlation • Shelf tag missing: An item is in distribution, but when
was 70 percent), which translates into an impact of about it is OOS, the shelf tag gets removed, thus making the
¼ of 1 percent shelf availability increase for every 1 percent OOS invisible to shoppers and store management.
increase in compliance. For the diapers category, the impact • Product hidden: When overflow product is hidden
was less, where the impact we observed was about 1/10 of 1 behind other SKUs, for instance, Cheer and Tide both
percent shelf availability increase for every 1 percent increase stocked on the shelf but some overflow Cheer product
in compliance. hidden behind Tide. This is often a symptom of wrong
shelf capacity and/or case pack sizes.
D. Data from other studies • Holes covered: If an empty facing is covered by another
Research by T Netherlands TU/e Retail Operations product then the situation of out of stock is not visible.
Group in dairy categories found planogram compliance to This is sometimes negligence but more often intended
be an important issue that requires a retail management behavior to hide out-of-stocks, either to display nicely
attention. Their evidence clearly linked the majority of stocked shelves to customers or hide out of stocks from
non-compliance with facing differences. They identified four store managers. In both cases it makes matters worse as
main drivers of non-compliance: 1) local store management, stocking errors are not identifiable.
2) a significant adoption time for changes, 3) different local
circumstances than assumed in the planograms (e.g., dairy) B. Tracking the Extent of Item Management
and 4) the lack of incentives from the headquarters. The Failures: A Study of Shelf Tag Removal
major consequence of a lack of planogram compliance proved Many retailers that use automation such as shelf label
to be a substantial loss of efficiency both in the marketing readers, do indeed “look for holes”. However in some retailers
strategy as in the operational executions, as such indicating it is accepted practice to “hide holes” by removing shelf labels
that planogram integrity is a serious issue if no product can be found. To measure the extent that this
practice occurs, a study was conducted in 24 stores at a major
3-7. Knowing Where Everything Should Be: Keeping U.S. retailer, examining a sample of approximately 100 CPG
the Shelves Straight items. Auditors visited the stores four-to-five days per week
While planograms are usually developed by over a three-to-six-week period. The auditors utilized an
headquarters, the actual maintenance of planograms in application running on a hand held, radio frequency (RF)
day-to-day operations is with the stores. Item management enabled, barcode reader. For each item in the study the
reflects the degree of maintenance of planograms, or in auditor had to read the shelf label and enter the quantity on
other words, the degree to which the products are in their the shelf. If there was no shelf label for that item, this was
assigned shelf space. In this section, we specifically examine indicated by a special key. A total of approximately 40,000
the impact that retail maintenance of shelf-tag accuracy, the observations were captured. Observations were eliminated in
number of OOS “holes” that get covered, the number of the cases where a store did not carry the product (it was not
SKUs with limited visibility or are otherwise hidden have on in distribution at that store), and thus a shelf label was not
the overall level of OOS. In a manual order situation, failures reported. Of the remaining observations, 7 percent indicated
in these merchandising practices will increase levels of OOS. the shelf label was not present. This finding suggests that
additional study of the effect of item management failures on
A. Common Failures of Item Management OOS was warranted.
As part of this research, we walked the shelves of
supermarkets around the world, and we found that products C. A Test of the Effect of Item Management
were often not in their assigned space. Often we found them Compliance
somewhere other than where the shelf-tag indicated they A European retailer participated in a study designed
would be, hidden behind other products, and found “holes” to unravel the cost of item mismanagement. The difference
covered by other items that would otherwise provide a clear between our proposed study and many others the retailer
signal that an item was out of stock. We synthesized our had previously considered was that most studies simply
observations and derived the following four common failures count holes in the shelf facings, but do not check thoroughly
of item management which must be addressed when trying whether products have the right tag, are hidden or are
to impose discipline in retail operations. covering OOS of other products.
• Wrong tag: Mismatch between the product assigned to
a shelf space as defined by a corresponding tag and the For this study, the retailer provided 20 stores that we
product actually occupying the space defined by the tag, divided into test and control store, being careful to match
for instance, Cheer sitting on the shelf space defined for each test store with a similar control store. To begin the
Tide. study, merchandisers went into each test store to examine

Chapter 3
the targeted categories (laundry, diapers, femcare, hair care), implementation of a disciplined item management process
and to check the degree to which items were in the space reduces OOS on average by four basis points.
they were assigned to according to the list of common item
management failures. The products were then “trued-up”, i.e., D. Conclusion:
an ideal situation was established by assigning products to Evidence clearly shows that item management failures
the right tag and taking hidden products as well as products in retailers are common, and our study shows that when
that hid out of stocks of other products off the shelves. It is these failures are addressed, that the impact on OOS levels
important to note that out of stocks were not replenished is substantial. In subsequent discussions with retailers, we
with backroom product. To the contrary, when the probed the degree that these failures were a matter of policy
merchandiser left the store, the shelves painfully displayed or practice. Here is what retail managers told us:
the out of stocks to both customers and store managers.
However, when the next delivery arrived, out of stocks were • Almost all retailers already have policies aimed at
of course eliminated. Merchandisers returned to each test eliminating these behaviors
store once a week to “true-up” any deviations. • Policies are not followed because of
o Discipline
The control stores were not “trued up.” Their state of o Less than 100 percent sell through of promotions
item management was checked at the beginning, at the o Inadequate shelf space for packout
middle and at the end of the three month test period, in o Too many items in assortment
order to allow a comparison between test and control stores.
Simple improvements is shelf tag accuracy, keeping
The results were striking, and are shown in the Figure product from being hidden behind other products and not
25 below. We measured OOS using the measure of lost units covering OOS holes will improve the accuracy of manual
and lost sales developed by Data Ventures. We found that ordering and lead to lower levels of OOS.
lost sales in the test stores where items were trued up were
on average only 6 percent compared to the control stores
were they were 10 percent. We found a particular strong
impact on promoted items. Overall, our research shows that
Figure 25
Credit: Data Ventures

Geting and Keeping
Lower Levels of OOS 4
Overview
Throughout Chapters 1-3, we have demonstrated 4-5. Bringing it all together, and looking forward.
several options available to retailers to reduce OOS. These Report conclusions
include keeping a focus on the items that account for the The future of OOS reduction
most lost sales, using new technology to measure and
identify the OOS and their patterns, getting better data to
make forecasts, keeping product information accurate (both 4-1. The ability to measure provides the ability to
product data and perpetual inventory), keeping shelves in focus
line with the planograms, keeping the items on the shelf in Measurement of OOS Leads to Efficient and
good order, and others. Throughout, we have emphasized the Effective Solutions
need for measurement of OOS to point to root causes that We have found that a major cause of failure to
will then provide direction for solutions systematically reduce and maintain lower OOS levels is not
focusing on the most critical root cause areas determined
In this chapter we present logical and systematic by analysis and continued measurement. It is crucial to have
approaches that retailers can use to address OOS. This the measurement capability to be able to point towards the
chapter is organized into the following five sections. dominant root causes, focus on the OOS that are the source
of the greatest loss, and continually feed back progress made
4-1. The Ability to Measure Provides the Ability to on OOS reduction.
Focus
Measurement of OOS leads to efficient and effective As we have shown in Chapter 2-1, such measurement
solutions can be manual, estimated by POS data, or tracked by
Measurement must identify the products with high perpetual inventory systems, and we have provided examples
OOS payoff and the stores with high OOS payoff of each approach. When manual measurement of OOS is
made through examination of a sample of items in a store,
4-2. Taking Action each item that is found to be OOS can be linked to its
Is it a store OOS or a shelf OOS? Is the root cause likely root cause through a systematic yet simple process of
product-related or store-related? checking each OOS against a hierarchical listing of known
The extent of implementation should be based on the OOS causes (i.e., begin with PI accuracy, if accurate then
resources that are available check to see it the item was on promotion, and so on). The
crucial objective for any measurement method is the need
to link the OOS to its root cause, which can then allow the
4-3. A Comprehensive Approach to Reducing OOS retailer to clearly see the dominant root causes and thus
Events, Duration, and Losses direct towards the proper solution. This manual approach
From forecasting to shelf replenishment and everything is appropriate for lower volume stores and for stores with
in-between slower moving items.
Data accuracy is the foundation
Can a retailer focus on everything? Alternatively, for large volume stores, and stores
dominated with faster moving items, newer technology
4-4. A Focus on the High Demand Items can be implemented to measure OOS on a continuing and
Constant high demand items efficient basis. We have also demonstrated in Chapter 2-1
Temporary high demand items that estimation through POS data can reveal OOS with great
accuracy. We recommend that retailers who have the resources

Chapter 4
and product movement characteristics to adopt this approach Based on this rationale, we recommend that retailers
for three reasons: create and systematically use some sort of POS-based OOS
measuring system, whether from a commercial vendor or
• First, OOS estimations made via this approach can be internally developed.
aggregated to show which products and which stores
tend to have higher levels of OOS. This provides the Measurement must identify the products with
ability to target and focus. high OOS payoff and the stores with high OOS
• Second, this approach provides estimates of multiple payoff
OOS attributes, which allows for the calculation of As discussed in section 2-1, not all OOS events have
multiple OOS rates. In particular, this provides the equal cost to the retailer, and similarly, not all efforts to
ability to estimate OOS loss rates, which increases the reduce OOS provide the same return on the resources
efficiency in targeting which OOS to address. invested. The second major recommendation is that retailers
• Third, the event frequency and duration can be linked seek to identify two dimensions of OOS loss: the products
to root causes, as we showed in Chapter 2-3. These that have the highest level of OOS and OOS-related loss,
identifying patterns provide a diagnostic approach and the stores that have the highest level of OOS and OOS-
to linking the root causes to solution areas. Figure related loss. When these dimensions are clearly identified,
26 provides an example of the way that typical OOS then an action plan can be put into place that can focus on
patterns can be linked to one or more of 17 retail these products and/or stores. This allows the retailer the
solution areas. choice to reduce the overall
OOS rate as efficiently as
Figure 26: Applying OOS Patterns to Likely Solution Areas possible, or it allows the retailer
to strategically select certain
SKUs, categories or stores.
4-2. Taking Action

Is it a store OOS or a shelf
OOS? Is the root cause
product-related or store-
related?
Our recommended
approach takes the following
three steps:
1. Measure and Assess
– High OOS risk products (fast
movers, or other
strategic SKUs)
– High OOS stores
– Shelf versus Store OOS
2. Solve high risk products with
Store OOS solutions
3. Solve high OOS stores with
Shelf OOS Solutions
Step 1: Measure
and Assess
In section 4-1 we focused
on how to measure OOS in
order to solve OOS. Once we
know where the OOS issues
rest, retailers need to form
an action plan based on the
Credit: Standard Analytics resources available. The action
Getting and Keeping Lower Levels of OOS
plan we recommend in this section is based on having such a The measurement system will identify both the best
measurement approach in place, because: practice stores as well as the high OOS loss stores. The
• If OOS problems are localized to stores then it is a store measurement will also identify the root causes that are
fix; and having the greatest impact on the OOS loss. The general
• If OOS problems are universal across stores then it is approach is to transfer the practices that address the
probably a product fix. dominant root causes from the best-practice stores to the
Addressing OOS in this manner will provide the high OOS loss stores.
greatest improvement for retailers relative to the resources
available. Figure 27: Plot of normal distribution
Step 2: Solve high risk OOS products with store

OOS solutions
Store OOS are generally product issues, i.e., the same
products are OOS across most stores. These are caused by
inaccurate product data, poor forecasting, insufficient delivery
cycles, or improper shelf-space allocation. Thus the solution
needs to begin with the items that generate the greatest
loss due to OOS. As we showed in section 2-2, the 80-20
principle applies store sales. In the specific example derived
from this research, only 360 items accounted for nearly 20
percent of the total OOS losses. As the specific items and
their impact are known, the place to start is to lower the Credit: Wikipedia.org
OOS losses on this relatively finite number of items.
The Extent of Implementation Can be Based on
Step 3: Solve high OOS stores with shelf
the Resources that are Available
OOS solutions
Shelf OOS are generally caused by issues related to Notably, the approach we recommend can be
specific stores, i.e., given similar store characteristics in terms implemented at the level of resources that the retailer has
of size, product mix and volume; some stores will have higher available. For example, if the measurement system identifies
levels of OOS than other stores. These differences are caused the 500 items that drive the greatest OOS loss, a retailer
by incorrect manual ordering, lack of planogram compliance, may not be able to address all 500 at one time if at all.
insufficient shelf replenishment processes and poor on-shelf Our recommendation is to work at the level of capacity to
execution. address a portion of the 500 items. If there are resources
to address 100 items, then deal with them. Addressing a
While we do not have specific analysis showing the portion will work to lower OOS losses. Moreover, OOS can
distribution of variance in OOS rates among stores in be addressed serially, such as addressing 50 at a time, solving
a chain, a helpful place to begin would be to assume a this group, and then moving to the next 50. For example,
normal distribution (as illustrated in Figure 27). If a normal if the measurement points to a root cause on insufficient
distribution was the case, then out of 100 stores, 15-17 shelf allocation that leads to OOS in a group of fast moving
would be “best practice” stores, demonstrating consistently items, then implementing demand-based planograms that
lower levels of OOS losses, 65-70 would be typical (within incorporate those items can substantially reduce OOS losses
one standard deviation of the mean OOS loss for all stores), across all stores.
and 15-17 would show OOS losses consistently below one
standard deviation of the mean OOS loss for all stores. Even A similar approach can be made when addressing stores
if a “normal” distribution was not the case for a certain group that have high levels of OOS relative to their peers. Selecting
of stores, there would still likely be a small group of best the 15 or 30 stores with the highest OOS losses and
practice stores that would set a standard for the rest of the addressing the replenishment and merchandising practices
group, and a small group of stores that would have larger at those stores does not necessarily require new technology,
OOS related sales losses. but improved processes and training. This can also be done
serially, one store at a time, or in small groups of stores that
have similar OOS loss characteristics.

Chapter 4
The point of this recommendation is to show that any there is the additional challenge of keeping both shelf and
retailer can follow this approach and reduce losses from displays stocked properly.
OOS, regardless of the level of available resources. One
retailer may make OOS loss reduction a major initiative, The order suggests that while errors at the top may be
while another may only be able to address a very limited compounded down the line, they may also be overcome, or
number of products and/or stores. Both can benefit. at the least mitigated by superior execution near the bottom.
On the other hand, errors at the bottom create problems that
4-3. A Comprehensive Approach to Reducing OOS will more likely directly create OOS with little chance of
Events, Duration, and Losses mitigation.
One objective of this report has been to systematically
present a comprehensive approach to OOS, following a 4-4. Practical Steps to Address OOS of High Demand
logical sequence of root causes and solutions. Throughout Items
Chapter 3, we provided a sequential examination of OOS We demonstrated that high demand items create
causes and solutions, which followed the order presented in the largest lost sales due to OOS, and in this section we
Figure 1. Addressing any of the root causes will lead to lower specifically address approaches to retailers who wish to
levels of OOS, however, sustaining a low level of OOS losses focus on these high demand items. This section takes
requires that each of the major areas and sub-areas shown several of the approaches and outlines their usage in a more
on Figure 1 be addressed, because they build on each other specific manner as applied to high demand items. High
and they interact. While retailers should use the action steps demand items have two categories, constant high demand
presented in section 4-2, there is a related comprehensive and temporary high demand. Each is addressed somewhat
approach to solving OOS that starts with data accuracy and differently. Figure 29 provides a summary of the approach
takes a logical progression through in-store replenishment. described below.
It is this comprehensive view that needs to be understood
to systematically attenuate OOS, and maintain these lower A. Identify constant high-demand OOS items
levels on a permanent basis. – As demonstrated in section 2-2, a relatively small
number of SKUs are sold on a daily basis.
Figure 28 presents a grid that shows the dependencies – Primary emphasis is to eliminate OOS on constant high
between the major root cause areas, and it also provides an demand items since movement is known and forecast
order of flow of possible errors in the total replenishment. should be accurate.
To begin, the grid separates store OOS (which primarily • Address high-demand OOS through safety stocks
result from errors in ordering and delivery) from shelf OOS on these items, enhancing backroom to shelf
(which occur mostly because of operations issues). This processes and use of demand-based Planograms
distinction is also helpful because a different set of people (POGs).
and processes are involved in store ordering than in store • RFID can play a role to enhance backroom
operations, and it matches the recommended approach in processes.
section 4-2. – Secondary emphasis is to reduce OOS (duration)
• Improve methods of sorting (off the truck). Triage
For store OOS, data accuracy is the foundational sorting into immediate OOS, direct to shelf, direct
building block for other sustainable solutions. Product to backroom. RFID can enhance sorting triage
data accuracy, perpetual inventory accuracy, and POS data • Shorten replenishment cycle
accuracy can drive improvements in OOS losses, and they • Real time notification of OOS so cycle is not
form the foundation for other solutions. Forecast error missed (should be few instances due to constant
and delivery error also contribute to store OOS, so that demand nature)
by the time the product gets to the store it has substantial
opportunity to not be the correct amount of product to meet B. Identify temporary high-demand OOS items
demand. (typically caused by promotions; also caused by
external factors such as weather, seasonality,
For shelf based OOS, the product delivered (whether
etc.)
the correct amount or not) is ineffectively moved to shelf
– Define as “expected lift of ___ percent by the promotion”
because shelf capacity may not handle the packout or
(typically defined as 4-6 x average daily movement)
demand needs. There can also be inefficiencies in the process
– Primary emphasis is to reduce the duration of OOS
of replenishing the product to the shelf either as trucks are
of temporary high-demand items. These cannot be
unloaded or moved from backstock. For displayed product
Figure 28: Overall View of Retail Practices that Impacts OOS

Figure 28 Overall View of Retail Practices that Impact OOS
TYPE ROOT CAUSE AREA ISSUE DESCRIPTION Usable Solution
Store Based Data Accuracy Product data is inaccurate; perpetual • Fix product data through data
Out of Stocks inventory is inaccurate; POS data is synch
“Ordering” inaccurate. • Improve PI accuracy
• Review POS scanning practices
Forecast and Order Accuracy Sales forecast is understated where Add back measure of lost sales
OOS are unknown/not adjusted for • Due to OOS in the estimate
in sales history. It is overstated • Due to poor execution
where SKUs in history benefited • Due to Data Synch errors
from switching due to OOS from
other SKUs.
Demand forecast starts with the
errors unknown from the sales
forecast and attempts to estimate
true demand based on unscientific
judgment.
Order Quantity Demand forecast adjusted by • Enhance PI accuracy

inventory quantity which is • Follow CAO recommendation for
frequently inaccurate. safety stocks level
• Don’t hide product
• Use RFID to track cases and pallets
Replenishment Delivery cycle is too infrequent to • Adjust delivery cycle to meet most
match demand for fast moving “stressed” items
items; quantity is different from • Monitor delivery frequency and
order; delivery arrives late timing
Capacity (Time Supply) Inadequate shelf capacity to meet • Demand-based planograms

frequent peak demand needs. Few • Safety stocks
systems calculate for their regular
peak demands.
Shelf Based Capacity (Packout) and Multiple of case pack required in • Improved packout practices
Out of Stocks Planogram Compliance shelf capacity for efficient • Adjust case pack sizes
“Operations” operations that avoids partial cases • Enhance planogram compliance
not fitting in the dedicated shelf
space. Good packout should avoid
product being hidden behind
adjacent product or in the
backroom. Packout can be poorly
planogrammed or maintained in the
perpetual inventory system.
Shelf Replenishment Inefficient and ineffective • Backstock to shelf process

movement of product from the • Sorting
backroom to the shelf, or from the • Notification of OOS
truck as it is unloaded to the shelf or • Shelf tags
into backstock.
© 2007 Th mas Gruen and Daniel Corsten. Page 45

Chapter 4
completely avoided, but the retailers can control the OOS in these items will have the greatest effect on
damage by shortening the duration. reducing lost sales.
• Can be reduced with improved backroom sorting • Measure OOS using the new technology that identifies
and backroom to shelf processes; OOS from POS data. The data is already being
• Can be reduced by improved notification of low collected, thus the measurement provides an additional
or OOS levels or unanticipated demand, and by application for getting a return on invested capital
reduced ordering/delivery cycle time; and and labor that is already allocated to collecting this
• Can be reduced by re-forecasting demand for the data. Estimates from these algorithms have reached
week based on the first day of sales. reasonable levels of accuracy, matching the accuracy
– Secondary emphasis is to eliminate OOS on temporary levels of the error-prone and labor intensive manual
high-demand items. audits. Moreover, this measurement provides an estimate
• Can be eliminated by carrying backroom safety of the actual sales that were lost while the item was
stock; OOS. Vendors of these systems provide reports that
• Can be eliminated by including feedback of show where the greatest levels of OOS occur, and they
OOS duration and estimated lost sales from point to areas that need to be addressed. This allows
previous promotions when estimating promotional resources to be efficiently devoted to reducing OOS.
demand; and • Identify OOS patterns. Many OOS function in a
• Can be eliminated by coordinated ad/price behavioral pattern that suggests the likely root cause of
reduction decisions with ordering cycles. the OOS and points towards the obvious solution.
• Fix the on-hand inventory counts (so that perpetual
inventory matches actual
Figure 29: Solution Grid for High Demand Based OOS inventory).
Because the ability of using

the POS data estimation of
OOS is so central to effectively
addressing and lowering OOS,
the data that goes along with
the system needs to be accurate.
POS and inventory systems are
inexorably linked in retail. Our
research showed two particularly
strong effects on OOS levels,
perpetual inventory accuracy, and
data synchronization.
The POS OOS estimates also

provide us with the information
necessary to address two additional areas that we identified
4-5. Bringing It All Together and Looking Ahead as critical for reducing OOS.
In summary, we have presented many solutions to • POS estimated OOS data leads us to make demand-
solving OOS and made numerous recommendations. based planograms.
One major story that we need to reiterate after • POS estimated OOS data helps us to arrive at a measure
researching a variety of efforts aimed at reducing OOS of true demand which is sales (observed from POS)
is that the new technologies that have been developed to plus lost sales (unobserved, but new technologies help us
estimate OOS from POS data need to be adopted and estimate).
pursued by retailers. Retailers should use this technology to:
• Identify the 20 percent of the items that constitute the As this shows, OOS estimation from POS data is a
majority of sales and concurrent OOS. This is fairly central core technological competency required to efficiently
simple to do given that POS data already contains and effectively address OOS, particularly in high volume
this information. Concentrate on these items when stores with many items that sell multiple units daily.
addressing OOS for two reasons. First, most OOS sales
loss comes from faster moving items. Second, reducing
However, we have also provided several It is still clear to us that we are one generation of
recommendations for smaller volume stores, and those technology away from having a complete solution—a
with slower moving items. It is a matter of linking the technology where we can identify products, match up the PI
measurement to the root causes. Simple processes can be and run the types of reports and identification of patterns,
followed to link manual measurement of OOS to root causes. all as one integrated informational sequence. It’s a pretty big
In addition, our general approach to taking action works for step, but all the pieces that we have to talk about are just that,
smaller or larger volume stores, as simply a matter of scale. pieces, and we are fitting them together in a report, and it’s
both easy and difficult. Easy because we understand each of
The Future of OOS Reduction: We Are Closer, but the pieces pretty well, but also difficult because we know the
Still a Step Away from a Final Solution of solution is still out of our grasp. We can significantly alleviate
Functionally Zero OOS Levels the pain, but we really don’t have “the final cure.”
Much like the economics concept of “full employment”
not actually being defined as “zero unemployment,” there So, there is more work to be done, continuing to look
is a logical structural level of OOS that are too costly for ways to reduce OOS and satisfy shoppers. But we
to eliminate. Our best estimate of this level is probably believe by instituting the recommendations we’ve offered
individual to the retailer and would be determined by looking in this report a significant permanent reduction in OOS is
at the best stores at the end of the normal distribution curve. achievable and measurable. Resulting sales improvements,
Zero OOS is not a realistic or intelligent goal. Imagine you cost reductions and positive impact on shopper loyalty
had full visibility of items on the shelf (item-level RFID, to stores are very real. Following our method you can
smart shelf tags and POS system), then finding out what is determine the amount of reduction you want to pursue.
out of stock on the shelf would be simple. However, as our
solution grid shows, that even if this is perfect, OOS could OOS are no longer a zero sum game. For years retailers
still occur to varying or unexpected demand, and therefore it have traded shoppers disgruntled because of OOS back and
depends on adequate ordering to effectively address OOS. forth with no real net difference in the number of shoppers
for any given retailer. But now some retailers are stopping
Technological solutions have arrived to help us reduce the loss of disgruntled shoppers through aggressive OOS
OOS. On the POS side overall OOS estimation is accurate, reduction programs. Those that choose not to gain control
but there are some sloppy practices. On the inventory system, of OOS will continue to lose shoppers but not get new
we know what is supposed to be in the store, but we often shoppers from other retailers. A thorough analysis of OOS
don’t know where it is in the store. is the only way to determine what plans and resources should
be devoted to improvement.

Chapter 4
Appendix
Appendix 1: Item Velocity and Lost Sales
3

Appendix
Appendix
Appendix 2a: List of References

1. Anupindi, Ravi, Maqbool Dada and Sachin Gupta (1998) Estimation of Consumer Demand with Stock out based
Substitution: An Application to Vending Machine Products, Marketing Science, Vol.17, 406-23.
2. Bill C. Hardgrave, Matthew Waller and Robert Miller (2006): “RFID’s Impact on Out of Stocks: A Sales Velocity
Analysis”, Working Paper, University of Arkansas, RFID Research Center.
3. Broekmeulen, Rob A.C.M., Jan C. Fransoo, Karel H. van Donselaar and Tom van Woensel (2006): “Shelf Space Excesses
and Shortages in Grocery Retail Stores”, Working Paper, TU/e Retail Operations Group, Technical University of
Eindhoven, The Netherlands.
4. Campo, Katia. Els Gijbrechts and Patricia van Nisol (2000). “Towards understanding consumer response to stock-outs”,
Journal of Retailing, 76 (2) 219-242.
5. Coca Cola Retail Research Group (1996): “Where to Look for Incremental Sales: The Retail Problem of Out-of-Stock
Merchandise”, Anderson Consulting
6. Corsten, Daniel and Thomas W. Gruen (2005): “Desperately Seeking Shelf Availability: An Examination of the
Extent, the Causes, and the Efforts to Address Retail Out-of-Stocks,” International Journal of Retail & Distribution
Management, 31 (11/12), 605-617.
7. Corsten, Daniel and Thomas W. Gruen (2004): „Stock-outs cause Walk-outs“, Harvard Business Review, May 2004.
8. Curseu, Alina, Tom van Woensel, Jan C. Fransoo, Karel H. van Donselaar and Rob A.C.M. Broekmeulen (2006):
“Modelling Handling Operations in Grocery Retail Stores: An Empirical Analysis”, Working Paper, TU/e Retail
Operations Group, Technical University of Eindhoven, The Netherlands.
9. DeHoratius, Nicole and Ananth Raman (2003): “Building on foundations of sand?”, ECR Journal – International
Commerce Review, 3 (1) Spring, 62-69.
10. DeHoratius, Nicole, Adam J. Mersereau and Linus Schrage (2005): “Retail Inventory Management When Records Are
Inaccurate”, Working Paper, University of Chicago Graduate School of Business.
11. Fisher, Marshall (2004): “To you it’s a Store, to me it’s a Factory”, ECR Journal – International Commerce Review, 4 (2)
Winter, 9-18
12. Fitzsimons, Garvin .J. (2000): “Consumer Response to Stockouts”, Journal of Consumer Research, 27, 2, 249-266.
13. F leisch, Elgar and Christian Tellkamp (2005): “Inventory inaccuracy and supply chain performance: a simulation study
of a retail supply chain”, International Journal of Production Economics, 95 (3) 273-385
14. Gruen, Thomas W., Daniel Corsten and Sundar Bharadwaj (2002): “Retail Out-of-Stocks: A Worldwide examination of
Extent Causes and Consumer Responses.” Grocery Manufacturers of America.
15. Papakiriakopoulos, Dimitrios and Katerina Pramatari (2005): “Detecting products missing from the Shelf utilizing a
Rule-Based System” Working Paper, Athens University of Economics and Business.
16. Sloot, Laurens M., Peter C. Verhoef and Philip Hans Franses (2005): “The Impact of Brand Equity and the Hedonic
Level of Products on Consumer Stock-out Reactions”, Journal of Retailing 81 (1, 2005) 15–34.
17. van Donselaar, Karel H., Vishal Gaur, Tom van Woensel, Rob A.C.M. Broekmeulen, and Jan C. Fransoo (2006): “An
Empirical Study of Ordering Behavior of Retail Stores”, Working Paper, TU/e Retail Operations Group, Technical
University of Eindhoven, The Netherlands.

Appendix
Appendix 2b:
Listing of Acronyms and Terms used in This Report
Appendix 3
Estimating of the impact of OOS on lost sales.
The basic estimation of the lost sales due to OOS requires knowing three pieces of information. The first is the OOS
rate for the category or organization of interest. In our 2002 study, we found that to average 8.3 percent. The second is to
understand the sales losses due to various consumer reactions. For example, in our 2002 study, we found that manufacturers
on average lose 30 percent of the sales when consumers confront an OOS item (due to brand switching and not purchasing
at all). The final is the total sales of the organization or category. The following Exhibit shows how this can be calculated.
OOS Rate ______% Example:

x Avg OOS rate 8%
Category Avg X
Lost Sales ______% MFR Avg Loss 30%
x X
Total Category/ Category Sales $1B
Organization Sales $______ =
= Lost sales $24,000,000
Sales Lost to OOS $______
Note that this is a rudimentary estimate of the impact of OOS, but it is a good starting point. The issues addressed in
section II of Chapter 1 of this report demonstrate that the true cost of OOS is much greater than the simple estimate of lost
sales of a given item or category.
Appendix
Appendix 4
Day of Week and Time of Day Patterns
The first chart shows

the typical weekly pattern
of OOS identified using
estimates from POS data.
This pattern matches the
pattern shown in our 2002
study which relied on in-
store audits. The second
chart shows the time of
day when OOS begin. By
identifying the peaks (Friday
afternoon, Saturday noon
and Sunday late afternoon)
retail managers can create
process and deploy stocking
labor to reduce OOS.
Appendix 5a
Studies Included in this Report
The studies that have led to the findings in this report were conducted in 2005 and 2006. Several of these were
conducted by the research team, some were commissioned by the research team, and others were conducted by other
researchers and integrated into this study. The major studies included:
• USA study that established best practices for planogram compliance measurement and examination of impact on OOS
• European study examining correlates of OOS
• USA study that examined perpetual inventory accuracy and OOS levels
• USA study that examined the impact of promotional forecasting and OOS
• European study that examined merchandising and shelving impact on OOS
• USA study that examined the effect of personnel longevity and deployment on OOS
• USA study that examined the effect of sales velocity on OOS
• USA study that examined the effect of RFID on OOS

Appendix
Appendix 6
Compliance Measurement
Compliance really breaks down into three components – distribution, space and arrangement. To do the final calculation
we calculate each of those components into a total with each component being one third of the total.
DISTRIBUTION
For calculating distribution an item in distribution that is not supposed to be is an error. An item that is supposed to
be there that is not is an error. Divide by the total number of expected items in distribution. That number is then multiplied
by two and subtracted from 100 percent. This multiplication is done because it is nearly impossible to get an error below 50
percent since to do so would mean the store would be carrying very few items at all regardless of authorization by HQ. They
would simply have empty space in the category which does not realistically occur. So this effect must be considered.
SPACE
Space is calculated on the items with a known expected space. If the item has a different number of facings than
expected, it is an error. The calculation is the number of errors as a per cent of the total number of items with an expected
space.
ARRANGEMENT
There are 3 parts to arrangement.
1. Is the item on the expected shelf is measured by taking the total number of items expected in the planogram with errors
in their shelf location as a per cent of total # of expected items found in the planogram.
n Example: if there is a vertical arrangement of the Pantene Brand and the SKU from shelf 1 is switched with a SKU
on shelf 2, that would be two errors and divided by the number of SKUs in the planogram that have been found (i.e.
Distribution errors are neutralized by not counting the items missing in distribution and not counting the items in
distribution, but not on the planogram)
2. Is the brand in the expected order is measured by determining by shelf how many brands are not in their expected order
and taking that as a per cent of the total number of expected brand/shelves in the planogram.
n Example: if a planogram is vertically arranged by brand from top to bottom on 5 shelves, and the second brand in
the order is switched with the third brand in the order that is two errors per shelf for a total of 10 errors. If there
are 10 brands in total all having items on all 5 shelves, that is a possible 50 brand/shelves. So the score would be 10
errors divided by 50 possible.
3. Is the brand item within the expected order is measured by determining whether each brand/shelf has exactly the right
order. If it does not that is an error and is calculated as a percent of the total number of expected brand/shelves in the
planogram.
n Example: if the planogram calls for skus to be ordered A-B-C-D-E and they are ordered A-B-D-C-E, that is one
error. If there are 50 total brand/shelves as in example in #2, then the calculation would be one error of 50 possible.
In fact any order other than A-B-C-D-E would produce an error in this calculation. If item B is not in distribution
and the order is A-C-D-E, this is not an order error – it would be a distribution error.
Planogram Compliance Checking Process

Make 3 “passes” through each category recording different information on each pass on the report provided.
Pass 1: Distribution Space and Shelf # collection

Look at the shelf and check the planogram report for
n Is the item in distribution anywhere on the planogram?
1. If yes, checkmark the “Distrib?” column
2. If yes, write down the number of facings in the “Facings?” column
3. If yes, write down the shelf number (#1 should be the bottom) in the “Shelf?” column
4. If no, write it down including UPC, item number on the back page
Appendix
Notes:
* If the SKU has a shelf-tag but there is an empty hole above it, the SKU is still considered there, but make a note of the
out of stock in the left hand margin.
* If an item is not on the planogram but has product on the shelf and the shelf tag is gone or has “Discontinued” on it, the
item is still considered in distribution and should be written down as an extra item on the last page. Please make a note
of the “discontinued” indication
Pass 2: Brand Arrangement collection

Check the shelf against the planogram report (read shelf from left to right)
n For each shelf in the report:
1. Pay no attention to the order of ITEMS within each brand – this pass is just looking at BRAND order
2. Determine if the BRAND follows the BRAND that is before it on the report (excluding missing items)
a. If so, put a “yes” in the green subtotal line under the “Brand Order” column
b. If not put a “no” in the green subtotal line under the “Brand Order” column
3. Extra items in distribution would create a “no” if they are not in the Brand order listed on the planogram report
*Note that for “mirrored categories” begin check from the right side of the category, and go to the left. Do this on every
shelf. If some shelves are mirrored and some not, read left to right.
Pass 3: Item Arrangement collection

Check the shelf against the planogram report (read shelf from left to right)
n For each BRAND by SHELF in the report:
1. Pay no attention to the order of the BRANDS, - this pass is only looking at ITEM order
2. Ignore any extra items in distribution that are not on the report as if they do not exist
3. Ignore any items missing in distribution – treat them as if they are in the correct order (example if item “B” is missing
from a lineup of “A-B-C-D” but the lineup is “A-C-D”, it is correct)
4. Determine if the ALL of the ITEMS within each brand follow the ITEM within the same brand that is before it on the
report
a. If all items do follow the proper order within the brand, put a “yes” in the green subtotal line under the “Item Order?”
column
b. If any do not follow, put a “no” in the green subtotal line under the “Item Order?” column
c. If you can’t determine the answer, write “can’t tell” in the green subtotal line under the “Item Order?” column.
How Compliance Gets Computed

Note: all calculations are deviation from 100 percent. Thus all of the following percentage calculations are “1 – error
percent”.
Calculation 1: Distribution Compliance. Expected SKUs not in distribution PLUS Excess SKUs in distribution DIVIDED
BY Expected # of SKUs on POG. The resulting number is the Distribution Compliance Percentage. Logic: For any
POG there is a set of items expected to be present. Variance from this set (those not there and those in excess of the ones
supposed to be there), detracts from the implementation of the plan.
Sub-Calculation 1: To create a basis for remaining calculations, so as not to penalize other measures for distribution failing,
take the total Expected SKUs on POG MINUS the Expected SKUs that were not found. That number will be less than
or equal to the Expected SKUs on the POG, and is referred to as Expected # of SKUs in Distribution.
Calculation 2: Facing Compliance. Number of Facing Errors DIVIDED BY Expected # of SKUs in Distribution. Logic is
that for the numerator and the denominator we have eliminated the distribution compliance variance.
Arrangement Compliance Calculations (3-6):

Calculation 3: Right Shelf Compliance. Number of Wrong Shelf errors DIVIDED BY Expected # of SKUs in
Distribution.
Calculation 4: Brand Block Compliance. Number of Brand Block Errors DIVIDED BY the Number of Brand Blocks.

Appendix
Calculation 5: Brand Block Item Order Compliance. Number of Brand Blocks with Item Errors DIVIDED BY Number of
Brand Blocks.
Calculation 6: Total Arrangement Compliance. SIMPLE AVERAGE of Calculations 3, 4, and 5.
Logic for Arrangement Compliance: Planograms are established for shopper and stocker convenience as well as to
maintain/enhance brand equity and store sales. The shelf order of the brands presents a logic to the shopper in terms of
position (what brand is adjacent to the brand), logical price points, and ease of finding desired items. The POG for the
category tells a story. The degree to which the arrangement adheres to that story, the better the story is communicated.
The storyline can be disrupted through placement of the brand blocks on the wrong shelf, mis-ordering of the brand
blocks on the shelf, and mis-ordering of the items within the brand block.
Calculation 7: TOTAL Compliance: SIMPLE AVERAGE of Calculations 1, 2, and 6.
Best Practice Approach to Planogram Compliance Measurement

The effective measurement of POG compliance involves three basic components:
1. Distribution: to what degree do the items on the shelves match the items that are supposed to be on the shelf?
2. Space: to what degree does the space allocated to the item on the shelf match the planogram? This is operationalized as
the number of facings for each SKU.
3. Arrangement: to what degree does the arrangement of the items on the shelf match the arrangement specified on the
POG? There are three components to this:
• Item on the correct shelf
• Brands arranged in the correct order on the shelf
• Items within the brand arranged according to the POG
To make the measurements, auditors reviewed each category in each store once a week, using a form we provided that is
shown in Figure XX below. Detailed instructions that we provided to store auditors can be found in the appendix as well as
the process used for calculating POG compliance from the data collected by the auditors.
About the Authors
Thomas W. Gruen, Ph.D.
Thomas Gruen is a Professor of Marketing at the University of Colorado at Colorado Springs, where he has served on
the faculty since 2001. He was on the faculty of the Goizueta Business School at Emory University from 1996-2001. He
holds Ph.D., MS, and MBA degrees in Marketing from Indiana University’s Kelly School of Business. Before entering the
academic world, he worked as a retail trade association executive for ten years, managing advertising and publications for
membership-based organizations.

Tom has been studying out-of-stocks since 2000. From 2000-2002, Tom obtained a major research grant from the
Proctor & Gamble Corporation, and led the research project team that produced “Retail Out-of-Stocks: A Worldwide
Examination of Extent, Causes, and Consumer Responses.” Along with his research partner from the London Business
School, Dr. Daniel Corsten, he is conducted a follow-up study on solutions to retail out-of-stock problems that is this report.
His other research focuses on the management of customer relationships. Most of his recent research has focused on
retailer-manufacturer relationships in the fast-moving consumer goods industry. His current research interests include retail
out-of-stocks, category management, supplier-retailer scorecards, and customer-to-customer value creation. His research has
been widely published including the Harvard Business Review, Journal of Marketing, Journal of the Academy of Marketing
Science, Journal of Retailing, Journal of Business Research, Journal of Applied Psychology, and ECR Journal.
At UCCS, Tom teaches retail management, e-commerce, and marketing strategy to both MBAs and undergraduates.
While at Emory, Tom taught Customer Relationship Management and Marketing Research to MBA classes as well as in
executive education programs. In addition to the above responsibilities, he is a regular Visiting Professor at ESCP in Paris,
France, is a member of the editorial review board for the Journal of the Academy of Marketing Science.
On a personal note, Tom lives in Colorado Springs, Colorado, with his wife and three children. A complete vita and
additional information can be found at www.uccs.edu/tgruen.
Contact:
Thomas W. Gruen, Ph.D.
Professor of Marketing, College of Business
University of Colorado at Colorado Springs
Colorado Springs, Colorado 80919, USA
(719) 262-3335
[email protected]

About the Authors
Prof. Daniel Corsten
Daniel Corsten is Professor of Operations and Technology Management at IE Business School in Madrid and formerly
Visiting Associate Professor of Operations Management at London Business School. Before he taught for more 10 years at
the University of St. Gallen, Switzerland, as Assistant and Associate Professor and Vice-Director of University’s Institute
of Technology Management. He also visited the Wharton School at University of Pennsylvania, Philadelphia, INSEAD
Singapore and Fontainebleau and Bocconi Unversity, Milan.
At the heart of Daniel’s research are supply chain management and service operations in diverse industries such as
retailing, computer, automotive, software and logistics services. Currently, he is working on a multi-year research project
on retail out of stocks with some leading retailers. His research appeared in Journal of Marketing and Journal of Physical
Distribution and Logistics Management as well in managerial journals such as Harvard Business Review. He authored several
books on supply chain management (in German).
A leading academic in the Efficient Consumer Response movement, Daniel was Founding Editor of the ECR Journal
and President of the ECR Research Foundation (2000-2006). He co-authored award winning case studies on collaborative
performance management and sustainable supply chains and has won two teaching awards. He is a regular speaker at
international conferences and a consultant to Fortune 500 companies.
Daniel Corsten holds a MSc from the University of Cologne and a DBA/Ph.D. of the University of St. Gallen. He has
worked as International Marketing Manager at Bayer AG, Leverkusen and as Project Manager at Agfa AG, New Jersey.
Contact:
Prof. Dr. Daniel Corsten
Professor of Operations & Technology Management,
IE Business School,
María de Molina, 11,
28006 Madrid
[email protected]

To Retail Out-of-Stock Reduction: A Comprehensive Guide

Uploaded by

Copyright:

Available Formats

To Retail Out-of-Stock Reduction: A Comprehensive Guide

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

To Retail Out-of-Stock Reduction: A Comprehensive Guide

Uploaded by

Copyright:

Available Formats

A Comprehensive Guide

A research study conducted by:

Supporting Trade Associations:

Partner Academic Institutions

Contact for permissions: [email protected]; [email protected]

Chapter 1: Understanding the Total Cost of OOS............................................................................................... 7-12

Chapter 2: Understanding OOS Through Measurement.................................................................................. 13-23

Chapter 3: Lowering OOS Rates—Our Hypotheses and Related Studies....................................................... 25-40

Chapter 4: Getting and Keeping Lower Levels of OOS...................................................................................... 41-48

Gruen & Corsten 2008 

Steps of Sustained Lower Out-Of-Stocks

2) Measurement: Knowing Where to Focus

3a) Lowering OOS: 3b) Lowering OOS:

Forecast & Replenish-

4a) Approach to Lowering OOS: Assessment and Implementation

4b) Sustaining Lower OOS

The use of point of sale (POS) data is a viable

Gruen & Corsten 2008 iii

A Methodology to Get and Keep Low Levels of Out Final Conclusions

Gruen & Corsten 2008 

Overview and Objectives After spending three years studying approaches to

Gruen & Corsten 2008 

a typical week an item—if always available—would Guide to Using this Report

Gruen & Corsten 2008 

Figure 1: Overview of the Chapters of This Report

Steps of Sustained Lower Out-Of-Stocks

2) Measurement: Knowing Where to Focus

3a) Lowering OOS: 3b) Lowering OOS:

Forecast & Replenish-

4a) Approach to Lowering OOS: Assessment and Implementation

4b) Sustaining Lower OOS

where managers can methodically diagnose and assess their

Gruen & Corsten 2008 

Gruen & Corsten 2007 

and these service failures point to lowered customer Large Small

In our 2002 study, we identified that on average 25

Gruen & Corsten 2008 

shown in Figure 4. At minimum, the cost to manufacturers

RFID applications are focused on knowing when the

Figure 5: Shopper Cost Components and OOS Behavior

generalized to entire categories and require refinement to be

There are still some critical things we don’t know about

Gruen & Corsten 2008 11

Gruen & Corsten 2007 13

The first is a manual auditing method where periodic

Gruen & Corsten 2008 15

Figure 6: Two examples of OOS Estimation Using POS Data

Credit: Data Ventures

Differences in the duration

Gruen & Corsten 2008 17

C. Measuring OOS Using Perpetual Figure 9 – Another View of OOS Duration

Our analysis of product movement shows that in a typical

Looking at the chart from another perspective, even on

This same pattern is even more pronounced for stores

This general pattern shown here has long been recognized

Gruen & Corsten 2008 19

2-3. Viewing OOS Patterns

Figure 13: Promotional Velocity Underestimated

Credit: Standard Analytics

Gruen & Corsten 2008

Gruen & Corsten 2008

Gruen & Corsten 2008

Gruen & Corsten 2008

Gruen & Corsten 2008

Gruen & Corsten 2007

Gruen & Corsten 2008