Dynamic Pricing and Inventory Management for

Retail Store using Consumer Transactional Behavior

Prabhakar Kadam
Information technology
Sardar Patel Institute of Technology
Nikhil Dalvi
Information technology
Sardar Patel Institute of Technology
Abstract We consider a periodic inventory control model in
which retail stores face problem of load of products not selling on
inventory and customer interest management. To address the
problem of products not selling, the system adopts a strategy,
under which it calculates load factor for each product using the
sales and current inventory data and suggest to create a
combination of highest and lowest selling product in a
subcategory to sell out the low selling product with profit. We
provide a solution for dynamic pricing of products as
combination. Our system removes the manual analysis part of
retail chains and automates the process. To address the problem
of customer interest management, system uses the transactional
behavior of consumer and provides the customer with deals or
offers by analyzing the trends in the history of consumer
transactions. For example, Generally customers of retail stores
are notified of all the offers through messages or emails which
amounts a lot of notifications and so remains unattended which
costs the firm a lot of money. Instead of this we can analyze the
transactional behavior of the customer and suggest only with the
offers that the customer might be interested in. In this way
notifications get reduced so there is very low chance that our
messages go unattended. It has a very high probability of
customer visiting back at store to avail the offer because it is
derived out of trends in customer behavior of purchasing
products.
Keywords (Load Factor, product category, inventory level,
probable customer)

Introduction

In todays highly unstable and competitive market

retail industry faces a lot of problems in inventory
management. New products are launched every day in
different categories of products. Customers demand for
newly launched product makes it very important to
clear older stocks [8]. Many times product sales fails
due to many reasons like over pricing, bad quality, lack
of advertising etc. Especially in food retail store
inventory management becomes highly tedious because
food products expire in short period, and is a total loss.
But above given reasons cannot be an excuse for retail
stores because there is a lot of money at stake. In any
situation you have to sell out the product and make
space for new product in inventory [4]. To tackle this

Vijay Patil
Information technology
Sardar Patel Institute of Technology
Prof. Swapnali Kurhade
Information technology
Sardar Patel Institute of Technology

problem we have derived an algorithm which

calculates load factor of products in a product
subcategory. Load factor is a term used to express load
of a product on inventory. Product selling lowest has
the highest load factor and the product selling highest
has the lowest load factor which implies that the
products with low load factor values are a load on
inventory and need to be cleared. Load factor for each
product can be calculated from sales figures and
current inventory level of the product. Suggestions
regarding the lowest and highest selling product can
then be given to the store manager who can work out
on how to sell out the lowest selling product by making
combo offers or any other way [3]. It may seem that if
butter is lowest selling and shoes as highest selling then
do we suggest to sell butter with shoes? No, algorithm
only compares load factors of products within same
category. For example within food products category if
some butter is lowest selling and some bread is highest
selling then algorithm suggests bread and butter. A
question that arises is that do we only care about the
one lowest selling product of that subcategory? No, it
is totally dependent on the software developer to
suggest how many products he wants to suggest at each
iteration he runs the algorithm. It is also obvious that
once a lowest selling product will not be lowest selling
in other iteration if inventory manager is smart enough
to sell it, and order low quantity next time or drop it
from inventory. Sometimes it may occur that two or
more products may come out with same load factor
values. In that case the algorithm can be extended to
find the per product profit that seller can earn. For
example, if two products X and Y come out to be low
selling with same load factor values then if profit after
selling one unit of X is greater than after selling one
unit of Y then it is always advisable to give preference
to X over Y. From business point of view it is logical to
earn more profit selling low quantity of a product
rather than to earn low profit by selling high quantity
of other product. There are many inventory
management algorithms used for example Apriori
algorithm. It is an algorithm for frequent item set
mining and association rule learning over transactional
databases. In this low selling products are missed out.
To overcome these drawbacks this algorithm works
efficiently. Still a major problem that rises is how do
we sell out the products suggested? To tackle this
problem we have designed a algorithm that provides

with an optimized value of selling price for the

combination of low and high selling product. We use
the comparison of load factor values of the two
products to come out with the selling price for the
combination. Mathematical model for the said
algorithms is explained in next heading.
In addition, we investigate the impact of offer
reminders that are used to let know the customers about
current offers and deals on various products. In the face
of growing number of products in a retail store there
are many products that have offers running at same
time. Messages or mails arrive too frequently which are
totally ignored. These reminders can be rather used
efficiently to increase sales by providing the customer
only the offers that he/she might be interested in. For
example, I go to a store only to buy large cold drinks
bottles, deodorants, maggie like stuff so sending me
offers related to vegetables, rice, oil is totally waste. In
our system we suggest to use the transactional history
of customer and find trends and shopping behaviour of
the customer and use it to suggest products and offers
[7]. In this way it is highly probable that customer
might not ignore the notifications and would visit the
store to avail the offer. Customer centric suggestions
can be generated in two ways
A. In this method we analyse the products that are
having low sales i.e. having high load factor in that
category of products. After finding the low selling
products we find high selling products in that same
category and then we find out the customers who
bought that product. These customers are our target
customers called probable customers. We can suggest
these customers the offers related that product because
these customers buy the products that are highly selling
in this category and might buy the other ones too [2].
B. In this method we analyse the customer by his
transactions. For a customer we find out the products
that he buys the most. After that we find the category
of those products and then find the lowest selling
product in that same category by comparing the load
factor values of the products. It is highly probable that
the customer might buy the low selling products from
the category of products he/she buys the most [6].

have focused highly on low selling and high selling

products and the way to find it by calculating load
factor. Also we have contributed towards a finding a
new way of notifying the customers about stores deals
and offers about the products they might be really
interested in, by analysing the transactional behaviour
of a customer.:
.
II. MODEL
We consider a inventory system of a retail store firm
that has a generalized hierarchical structure within its
products as shown in Figure 1.

Fig 1
Load factor of products is calculated to further suggest
the low and high selling products within a subcategory
using the following algorithm.
Algorithm 1: Step 1: Load factor(L) is defined as the
ratio of remaining inventory value to the total
inventory. Total inventory is the sum of remaining
inventory an sold inventory of product.
L= where R= remaining inventory
I= total inventory

Using any of the above two methods we can provide

suggestions to the customers that are highly probable to
work out and thus result in high sales. Looking at
current scenario of Indian multi brand retail inventory
management is mostly done manually except in some
very large retail chains like wall mart. Manual analysis
and management is not an efficient way and consumes
a lot of work force and time and can even lead to
errors. So it is very important to use automated
software systems for analysis of sales and remaining
stocks in inventory.

Step 2: We select the product with the highest load

factor(L) from the inventory. Higher the value of L,
higher is the load on inventory. Our aim is to minimize
load on inventory. So we will try to couple items with
higher and lower load factor.

To conclude this, we summarize our main contribution

as overcoming the drawbacks of currently available
algorithms and providing solutions for effective
management of inventory and optimize profit of retail
industry by providing dynamic pricing solutions. We

b. Parents for the selected product must vary to avoid

the conflict of repetition of the same product.

The selection of the product to couple with the current

selected product can be done by following rules:
a. Products parent must have same parent as the
selected products parent.

c. Product should have the lowest load factors among

the products satisfying above two criterias.

For Example Let the following tuple show the sample

of the products available. Each value in the tuple shows
the different products and the combination of tuple
shows the subcategory of the products. Here the
subcategory tuples work as the child for the category
tuple.

Conditions: For load factor(L) >0.5 i.e. for product

selling below average

{
{a1,a2,a3,a4,a5,a6},
{c1,c2,c3,c4,c5,c6} }

Let the combined price of product1 and product2 after

dynamic pricing be z2. It will be depend upon the load
factor(L).

{b1,b2,b3,b4,b5,b6},

Let if the product b2 have the highest load factor then

to find the appropriate product to pair it with, we
search the product with lowest load factor in the tuples
other than tuple containing b2 i.e. even if any product
say b6 have lowest load factor then also it is not
allowed to avoid the repetition of the product class in
the combination .The all remaining products from
above sample are allowed to be form combinations [5]
Step 3: It is possible that some products might have
same load factor. In that case we will calculate profit
factor (p).For calculating the profit factor we consider
the absolute cost of the product at that moment which
can be calculated as follows:
C=C + T +Ic where C= Cost Price
T= Transportation cost
Ic= Inventory cost
Profit =S-C where S= Sales price
and Profit Factor P=(S-C)/C
When the two products will have the same load factor
then we will decide the combination by combing the
selected product with the product having high profit
factor among the products having same load factor.
Step 4: If the load factor as well as the profit factors
are same then we go with product with highest
procurement cost.
After selection of products we need to decide the
selling price for the combination of product. Dynamic
pricing can be efficiently done using following
algorithm.
Algorithm 2: Suppose we have following two products
whose combined costing we need to find
Product selling price procurement cost sold products
remaining inventory
Product1 s1 c1 sn1 in 2
Product2 s2 c2 sn2 in 2
Let the value of load factor for Products1 and Product2
are L1 and L2 respectively.
L1=11+1 and L2=22+2
The price for the products decreases from the sales cost
of the product to the procurement cost of the product as
the load factor starts increasing i.e. costing is inversely
proportional to the load factor [1]
Cost 1

Let the combined price of product1 and product2

before dynamic pricing be z1.
z1=s1+s2

z2= s1+c2(1+(s2-c2)(1-2)); for L1>L2 & L1<0.5

s2+c1(1+(s1-c1)(1-1)); for L1<L2 & L2<0.5
s1+s2; for L1==L2
Using above algorithm automatic dynamic pricing is
achieved efficiently.
III. CONCLUSION
This paper studies the Inventory management model of
retail store under fluctuating inventory level using
consumer transactional behavior. To effectively hedge
against inventory loss risk and load of low selling
products on inventory system suggests the inventory
managers of the low and high selling products of each
category by calculating the load factor for each product
and comparing them within themselves. Our study
investigates the crucial role of regular clearance of
stock in inventory to make space for new inventory and
keep the money flowing. This paper also provides an
efficient solution for dynamic pricing of the
combination of products. One focus of this paper is to
study the impact of proper use of offer reminders for
customers in retail industry. It focuses on customer
centric suggestions by analyzing the transactions of
customer and suggesting list of products from low
selling range that are highly probable that the customer
might buy. We stress on the use of software systems by
retail firms to automate the process of inventory
management and optimize profit.
ACKNOWLEDGMENT
We have immense pleasure in presenting the report for
our paper entitled Inventory Management of Retail
Store under Fluctuating Inventory Level using
Consumer Transactional Behavior. We would like to
take this opportunity to express our gratitude to a
number of people who have been sources of help and
encouragement during the course of this project.
We are very grateful and indebted to our project mentor
(Mrs.) Swapnali Kurhade, for providing her enduring
patience, guidance and invaluable suggestions. She was
the one who never let our morale down and always
supported us through our thick and thin. She is a
constant source of inspiration for us and took utmost
interest in our project. We would also like to thank all
the Staff Members of Sardar Patel Institute of
Technology for their valuable cooperation and
permitting us to work in the I.T. Lab.
REFERENCES

[1] Victor F. Araman, Ren Caldentey , "Dynamic

Pricing for Nonperishable Products with Demand
Learning", Operations Reaserch, Volume 57 Issue 5,
1169-1188, 2009
[2] Matt Adams, " Predictive Analytics for the Retail
Industry", Microsoft Technical Review, 1-25, 2008
[3] Vidya Mani, Saravanan Kesavan, Jayashankar M.
Swaminathan, Estimating the impact of understaffing
on sales and profitability in retail stores,
Server1.tepper.cmu.edu, 2011
[4] Kanna Miwa, Soemon Takakuwa, Simulation
Modeling And Analysis For In-Store Merchandizing Of
Retail Stores With Enhanced Information Technology,
Winter Simulation Conference, 2008
[5] Abbas, W.F., Ahmad, N.D., Binti Zaini, N.,
Discovering Purchasing Pattern of Sport Items Using
Market Basket Analysis, Advanced Computer Science
Applications and Technologies (ACSAT) IEEE, 120126, 2013
[6] Churinske, L.M., Lee, T., Linhares, N., Turbat, S.,
Intelligent marketing strategies for the Hard Rock
Hotel and Casino, Las Vegas; market segmentation and
customer analysis, Systems and Information
Engineering Design Symposium (SIEDS), 13-18, 2010
[7] Dr.B.Rambabu, Dr.G.Malyadri, A Study on
Inventory Management at Amara Raja Electronics Ltd,
Tirupati, Global Journal for Reaserch Analysis
Volume 3 Issue 1, 44-47, 2014
[8] Dr. Prafulla Sudame, Brijesh Sivathanu,
Challenges Affecting the Organized Retail Sector,
Arth Prabhand: A Journal of Economics and
Management Volume 2 Issue 8, 2013