ISIR Symposium 2016 Session Reports

19th International Symposium on Inventories
August 22-26, 2016 Hotel Hilton Budapest, Hungary
Session chairs report
AUGUST 22, 2016 distributions. Examples are used to illustrate the

Opening Plenary Session advantages and difficulties of each approach.
Maximiliano Udenio, Kai Hoberg and Jan

by Ou Tang
Fransoo: Inventory dynamics in the financial cri-
This plenary session included two interesting pa- sis: an empirical analysis of firm responsiveness
pers. and its effects on fincial performance
Rommert Dekker: Some new avenues for spare In this presentation, the authors study how firms
parts demand forecasting response when facing a drastic change of indus-
trial environment. Using the firm-level data of
1278 public US manufacturing firms, Maximiliano
Udenio and his colleagues first conclude that
firms often take as long as 6 quarters to adapt to
the situation. However, firms responsiveness var-
ies, for instance firms in chemical and electronic
sector often take the longest time. They also ex-
plore the factors which have impacts on the re-
sponsiveness of inventory policies, and these fac-
tors include liquidity, gross margins and inventory
size. The study confirms that managing invento-
ries is indeed critical in the crisis period.
In this presentation, Rommert Dekker presents the

challenges of spare parts forecasting, due to the
low volume, intermittent and erratic patterns. In
addition to the historical data, he suggests to
include additional information from other
sources. First, installed-base data can be used to
incorporate preventative maintenance and ser-
vice contracts can be used for improving inven-
tory control. Secondly, condition monitoring
method can be used to foresee the needs of
spare parts, even though there is a risk of false
alarm. Finally extreme theory is proposed to im-
prove the estimation of tails of empirical demand
19th International Symposium on Inventories Session Reports
August 22-26, 2016 Hotel Hilton Budapest
Opening Plenary Session 2 nance decisions jointly with spare-parts inventory

replenishment decisions should offer economic
by Robert W. Grubbstrm advantages, compared to separate individually
optimised policies for maintenance and replen-
In the second part of the Plenary Lectures Mon- ishment. The model was formulated in terms of a
day Morning, three papers were presented. Markov process for breakdowns with long-run
average costs as the objective function to be
The session started off with a paper entitled Re-
minimised.
plenishment and ordering decisions for fresh pro-
duce with backroom effect co-authored by Lin The results from this analysis indicated that the
Li, School of Business, Eastern China University of popular reorder point (s,S) policy could be far
Science and Technology, Shanghai, and the from economical, if the interactive effects from
President of ISIR, Ou Tang, Linkping Institute of maintenance and replenishment decisions for
Technology. This paper dealt with a fresh-food multiple-component systems were not taken into
inventory problem, when inventory is split up into account appropriately. The savings from joint
a front display area from which customers pick considerations could indeed be substantial.
their products, and a back room from which re-
plenishments are made. These two storage facili-
ties have typically different deterioration rates,
the back room having the slower rate, and cus-
tomers will choose according to a Last-In-First-Out
(LIFO) principle, attempting to maximise the
freshness of their individual purchase. A General-
ised Reduced Gradient algorithm was used for
solving the mathematical model. The results of
this study indicate that there is definitely a profit
to be gained by taking the gap between dete-
rioration rates together with the LIFO behaviour
into consideration when choosing the appropri-
ate inventory replenishment policy for the front The Secretary-General of ISIR, Krisztina Demeter,
shelve. presented the third paper Overview of ISIR Spe-
cial Issues Key topics identified by text mining,
co-authored by Andrea K, both authors affili-
ated with the Corvinus University, Budapest. With
a text-mining method, they examined the occur-
rences of topics (bags of words), which had
been given most attention through the recent
eleven ISIR symposia, by examining all texts from
The second paper presented, Joint condition-

based maintenance and inventory optimization
was written by the Incoming President of ISIR,
Ruud Teunter, University of Groningen. The basic
argument underlying the model formulation in
this paper is that taking condition-based mainte-
2
articles published in the Proceedings Issues of the This figure is one of several examples of the au-
International Journal of Production Economics. thors findings, in this case showing a strong in-
Text-mining involves a semi-automated process crease over time of terms involving stochastic
for finding structures (knowledge) in unstructured vs terms involving deterministic. Other conclu-
data (which encompass 85-90% of all company sions include finding five clear clusters (SCM,
data). forecasting, time, optimization, inventory mana-
gement), terms with increasing importance over
time (logistics, standardization, spare part mana-
gement, profit), and that the ISIR section topics
were relatively well-balanced.
The three presentations were most interesting,

clearly showing the synergetic effects from more
global considerations compared to sub-
optimisation (two first papers) and the opportu-
nity to gain knowledge from unstructured data
by a syntactic methodology (third paper).
Economics of Inventories
AUGUST 22, 2016 EC1 bles for industry (services/manufacturing) and

technological content. The findings by and large
by Geert Langenus indicate that both groups of certified firms exhibit
higher financial performance in comparison to
In the EC1 session that was chaired by Geert the group of non-certified firms. The correlation
Langenus (National bank of Belgium) two very between the ISO 9000 certification and firm per-
interesting papers dealing with the macroeco- formance also seems to be of a lasting nature as
nomic aspects of inventories were presented. it endures in the post-certification period.
Efrosini Siougle (Department of Informatics, Ath- Session participants generally welcomed the re-
ens University of Economics and Business) pre- search and speculated about the reasons for
sented a joint paper with Claire Economidou and these findings. They wondered, in particular,
Sofia Dimelis that looks into the link between ISO about the direction of causality: are efficient
9001 certification and firm performance. To this firms more prone to seek ISO 9001 certification or
end the researchers have compiled an extensive are the certification requirements pushing firms to
and original data set covering all firms that were behave in a more efficient manner?
listed on the Athens stock exchange from 1992 to
2013. Using a difference-in-difference approach The second presentation by Attila Chikn and
they analyse in particular how indicators of firm Zsolt Matyusz (both Corvinus University of Buda-
performance, including the return on equity and pest) specifically looked into the industry charac-
on assets, the profit margin, as well as indicators teristics of inventory investment in a broad range
related to efficiency in inventories (e.g. inventory of developed economies using OECD data. This
turnover), differ between three groups of com- is part of ongoing work on a book chapter with
panies: those that were ISO 9001 certified other colleagues of Corvinus University and the
throughout the period, those that obtained certi- Institute of Economics of the Hungarian Acade-
fication during the period and those that are not my of Sciences). They start off from the observa-
certified. The regressions contain control varia- tion that, from a macroeconomic perspective,
3
the average inventory investment as a percent- for each planning period, seasonality of custom-
age of GDP is trending down for a long period er demand which can exceed production ca-
already and analysed to which extent changes pabilities, limited storage space for multiple
in sectoral or industry characteristics play a role in stock-piling purposes, and thriving efforts for min-
this development. Given the length of the time imizing employed capital in physical inventories.
series that start as early as 1970 for most coun-
Hedging is widely used in companies trading
tries, the researchers are able to map how econ-
physically substantial amount of commodities in
omies change over time. They specifically use
order to mitigate risk over an investment. The role
cluster analysis to define common patterns and
of inventories is to fill the gap between geo-
identify different groups of countries that exhibit
graphical and timely disruptions of the customer
the same trends. While they acknowledge that
supply process. The dynamics of inventory levels
certain country-specific characteristics matter
at the magnitude of bulk commodities produce
(e.g. the oil industry in Norway), they argue that
significant price exposure, which requires turning
the declining level of average inventories can
to methods widely discussed in scientific litera-
indeed be traced back to a large extent to shifts
ture.
in the sectoral composition of output. In this con-
nection, the decreasing importance of different After examining the relevant literature review
manufacturing industries in many developed they analyzed different live cases when stock-
countries seems to be key. However, they also piling proved to be financially beneficial if price
show that this is not necessarily a linear trend as exposure on inventories was ruled out. The results
results from the cluster analysis differ between the can be quantified, different scenarios were pre-
two sub-periods considered (1970-1993 / 1994 sented as the basis of inventory management
2013). decisions. The paper concluded in endorsing the
active usage of hedging in order to support mul-
Session participants generally found the cluster-
ti-period inventory decision making for higher
ing approach advocated by the authors very
profit levels.
interesting. They agreed that shifts in the compo-
sition of output should be taken into account Horizontal coordination under information distor-
when analysing or interpreting developments in tion by Silvia Valeria Padilla Tinoco, Robert Boute
inventories. The authors findings also point to
further research avenues that may aim at deep- A two-company information sharing game was
ening our understanding of the importance of studied under horizontal coordination, using a
these sectoral shifts in the economy. can order policy. The authors evaluated: (1) the
influence of distorting information regarding de-
(A third presentation that had initially been mand rate on individual expected profits, (2)
scheduled was dropped from the agenda due whether information distortion could be eliminat-
to the unavailability of the presenter.). ed by selecting an adequate cost/gain alloca-
tion mechanism.
AUGUST 23, 2016 EC2 An embedded Markov chain model was used to
obtain the exact number of orders triggered (or
Creating value with inventories in a commodity joined) and to obtain the steady-state inventory
supply chain levels. They found how the demand distortion
Peter Kalocsai and Peter Borbs presented an impacted both transportation and inventory
industrial application of commodity inventory costs of each individual company. By introducing
hedging. The question in focus was how deriva- a certain margin of demand error, the company
tive trading could create value in a commodity might be able to influence the proportion of the
(crude oil and its derivatives) industry with the costs/gains to be allocated. For the simple de-
assumptions of using multi-period and multi-plant mand allocation rule no equilibrium was found.
LP optimization for generating inventory outlook
4
Given that the expected profits strongly depend last 40 years. By means of a (log-log) regres-
on both the margin of demand error and the sion analysis the authors find that the effect
selected allocation mechanism, these infor- of change in the CCC creates a cash-flow
mation distortions might reduce the benefit or bullwhip effect further up the supply chain.
even prohibit information sharing in supply
Questions
chains. It was found that by allocating the real-
ized costs (in contrast to expected costs) of the Is there a clear win-win situation for both the
coalition, the incentives to distort the information supplier and the buyer?
of demand are reduced. The presenter argues that on average the
net effect on the analyzed data indicated
Taking stock of inventory investment in Bel- that this was the case.
gium by Geert Langenus and Rasa Basselier As a general suggestion it was pointed out
This paper aimed to improve the understand- that it is beneficial to squeeze the supplier,
ing of the movements in Belgian inventories, as but not too much. It would be a better ap-
reported by the National Accounts Institute proach to reduce inventories instead of in-
(NAI), and analyses, in particular, to what ex- creasing payable days.
tent a clear cyclical pattern could be dis-
An EOQ-based working capital minimization dis-
cerned. It was also assessed whether relevant
counted cash flow model with infinite production
information on stock movements could be
capacity
gauged from the high-frequency data, such
Thomas Yeung - Yuan Bian - David Lemoine
as survey data, that were available in a more
Nathalie Bostel-Dejax -Vincent Hovelaque - Jean-
timely manner than the national accounts.
Laurent Vivani
Finally, the authors explored the role of explicit
model forecasts of inventory investment in the The study contributes to the literature by
context of the National Bank of Belgiums analyzing the EOQ model and Working
(NBB) biannual macroeconomic projection Capital Requirements (WCR). As such, opti-
exercises and challenged the current techni- mizing physical and financial flows.
cal assumption that typically set all future con- The EOQ-discounted model considers two
tributions of inventory investment to GDP interest rate parameters (i.e., and )
growth to zero.
which account for delays and risks in credit
from the producer stand-point.
AUGUST 25, 2016 EC3 Within the Tactical Production Planning
(TPP), the authors assume one-to-one bill of
by S.Padilla Tinoco materials (i.e., one raw material is used to
produce a final item with no wastes).
The Cash Conversion Cycle: An Empirical Analy-
sis of its Impact on Supply Chain Performance Questions
Maximiliano Udenio - Shaunak Dabadghao The approach of including WCR is a good
idea, but there may be two issues: (i) the
The authors study the drivers behind the
planning period (turnover) in the rolling hori-
Cash Conversion Cycle (CCC) in a supplier-
zon (i.e., 1 year versus 3 months approach),
buyer relation based on a VMI.
and (ii) the fact that pipeline inventories and
Their main goal is to see whether increasing
receivables are financed by the WCR.
the CCC of the supplier (i.e., as the supplier
The presenter argued that the model is a
has more liquidity) would increase the CCC
real time evaluation of cash-in and cash-out
of the buyer.
flows (in every period). By means of the pa-
The empirical study is based on firm-level
rameter , the model considers delay in
data coming from the NAICS 31 33 of the
payments.
5
Inventory Management
AUGUST 22, 2016 IM 1 ti-echelon inventory systems with lost sales. The
paper proposed a Markov decision process for-
by Engin Topan mulation for the optimal inventory policy and
also an approximation for the base stock policy.
There were 3 talks in my session. All the talks were
None of the talks get a major criticism. The time
presented according to order announced in the
management was not perfect. The contents of
symposium program. There were around 25-30
the presentations do not perfectly fit. Yet presen-
people in the session. This was pretty much same
tation subjects were interesting and presented
during the 3 presentations. The audience was the
papers were mature, presenters were good. To
people who mainly follow the service logistics
my experience, it was one of the successful ses-
SERL and inventory management IM tracks.
sions in the symposium.
The first speaker was myself. Although now I am
working at University of Twente, the paper that I
present is a part of my work at Eindhoven Univer-
AUGUST 23, 2016 IM 2
sity of Technology. The presentation was about
using imperfect advance demand information by Refik Gll
(ADI) for the optimal control of inventories multi-
ple items for a setting in which it may be eco- The session had three interesting talks, and mod-
nomically unattractive to collect ADI for all items. eling and understanding the customer behavior
In the paper, we propose a model and a solution was the common theme in all the three talks.
method to select items for which ADI is collected
and to find the optimal ordering, and return poli- The first talk in this session, titled Consumers as
cy for each item minimizing average long-run inventory holders, was given by Hisao Fujimoto.
total cost. The presentation took 27 minutes, and In this talk, Hisao Fujimoto argued that the pur-
questions took 4 minutes. I got 3 questions from chasing behavior of Japanese customers have
the audience. They were about why information changed considerably over years, and hence
update is not considered, and whether the these changes should be taken into account in
method is really tractable for practical size prob- planning retail networks. The proposed consumer
lems. buying behavior model partition consumers with
respect to frequency and quantity: from frequent
The second speaker was Johannes Fichtinger centralized buyers to non-frequent decentralized
from Vienna University of Economics and Busi- buyers. For instance, a frequent and decentral-
ness. His talk was about segmentation strategies ized buyer would make purchases in smaller
of supply chain network and its benefit on safety quantities and from specialized mini-stores. Such
stock reduction. The presentation took 27 a classification of consumers may also suggest a
minutes, and questions took 3 minutes. The willingness-to-pay behavior of corresponding
speaker got 2 questions from the audience. The consumers. Hisao Fujimoto concluded the talk by
questions, which layer lead to a discussion, were suggesting that the proposed classification may
about the concerns about whether segmenta- lead to specialized inventory models.
tion might lead to loss of synergy in the supply
chain. The second talk, titled Modeling and analysis of
primary and secondary markets in a system with
The third speaker was Marco Bijvank from Univer- random unit costs was presented by Refik Gll
sity of Calgary. His talk was about analysis of mul- (co-authored by Ycel Grel). Refik Gll pre-
6
sented a model system where the customer de- ient return process. It turns out that the amount of
mand can be partitioned into two segments: a refund offered does not have a significant im-
primary and a secondary market, and argued pact on the purchase intention or trust.
that these kinds of systems are observed, for in-
stance, in technology intensive products or ser-
vices where the primary market, being more loy-
al, is generally not too sensitive to the pricing of
AUGUST 23, 2016 IM 3
the product or service. Also, as the secondary
by Judit Olh
market customers are more sensitive to the sales
price, a demand model, where the customer There were two presentations in this session.
demand explicitly depends on the selling price
seems to be more appropriate. Refik Gll de- The Integrated Approach of Order Picking Sys-
scribed a procedure for obtaining the initial tem in Warehouse Management: a Case Study
quantity to order and the optimal sales price for by Judit Olh, Jlia Pauluk and Jzsef Popp, Uni-
the secondary market customers. versity of Debrecen.
Order Picking System is a warehouse manage-

ment process module which helps to put to-
gether the range of goods from warehouse
stocks based on individual orders. The fundamen-
tal problem of this enterprise is to identify the or-
der that the order picking employee should fol-
low when accessing all locations in the shortest
possible time and route. In order to shorten the
picking travel time, we initiated the change of
the order of products on the order picking list.
The order picking list based on the numerical in-
crease of the original ascending CAI (Code Arti-
The last paper of the session, Effect of online re- cle International) identifier was switched to the
tailers return policy on purchase intention was order picking list based on the alphabetical order
presented by Jing Zhu (co-authored with Fy of storage locations. It was concluded during the
Xiaorong and Xie Qinghong). The authors, using research that if storage location involvement is
data collected from a questionnaire, claim that below 10-15, the amount of time to be saved is
there is strong correlation between trust (affect- minimal (0-4%). However, even up to 11-13% time
ed by the online retailers return policy) and can be saved in the case of storage location
online purchase intention. One of the important involvement of 70-90. It was concluded that
conclusions of the research is that, an online re- preparation has to be done in as many storage
tailer can increase the sales by offering a suffi- locations as possible during commissioning. The
ciently long return window along with a conven- higher the number of goods is, the more time
7
can be saved. Consequently, preparation per- development of our sotfware, it is shown that
formed based on the desired new commission list production and inventory control IoT system can
is more effective at higher storage location in- automatically control inventory shortage. That
volveme nt. Employees performing Order Picking may decrease the number of production man-
System can obtain an ideal route with a order agers from production lines whose salaries are
picking list based on the ascending alphabetical more expensive than workers and it is hard to
order of storage location names, since these standardize managers skills.
employees enter a room only if the good in the
given room is on the order picking list.
Introduction of IoT system to real time produc- AUGUST 23, 2016 IM 4

tion change with shop inventory management
by Hiroaki Fujikawa and Kouhei Kominato, Tokio by Krisztina Demeter
University of Science
There were three presentations in the session,
The fresh food commodities like sandwiches or each of them self-managed, since the session
lunch boxes sold in convenience stores must be chair unfortunately was not present. Approxi-
replenished continously in order not to face the mately 10-15 people visited the session, which
shortage of commodities on the shelf. In general, took place on Tuesday afternoon.
most of the convenience stores keep their com-
Martin Albrecht (Paul Hartmann AG, Heidenheim,
modities in stock all the time. Then the factory of
Germany) was the first speaker, dealing with the
such commodities has to know the inventory lev-
issue of spare part inventories. He presented a
el on the stores every moment and produce ap-
practical case study, where a heuristics was de-
propriate item and appropriate quantity of the
veloped using item-based approximations for the
commodities. Production schedule of fresh food
order-based back-orders. The heuristics is very
commodities must be prepared in advance and
promising, since it does not need very detailed
executed under the control of production man-
data but provides close to optimal solution. The
ager who sould graps progress of each com-
solution does not take much time. Since the heu-
modities and workers conditions concurrently
ristics was tested on independent and identically
and make a decision how to share the excessive
distributed demand, the next step is to test it with
works among workers. It needs real time progress
real-world data, where this condition is rarely
monitoring. We introduce monitor camera in-
hold.
stalled in factory line-side to get the image data
which is taken all day long to fint changeover Florian Kaiser (Passau University, Germany) pre-
instructions to workers. The image data from line- sented a research carried out jointly with his col-
side monitor camera could be analyzed and league, Robert Obermaier. The purpose of the
grasped present progress conditions of each presentation is to investigate the causal relation-
worker and each commodity by the developed ship between inventory management and firm
sotfware which encompass both image data performance. Although previous research con-
and POS data from convenience stores. A de- sidered the impact of inventory management on
veloped software is examined under the test firm performance, there is also argument for the
production line. Where the production items reverse relationship, where the firm performance
should be changed in accordance with the de- affects the level of investment into inventories.
mand fluctuated condition. Production items The presented empirical study is based on data
have changed successfully to cover the ad-hoc from the German manufacturing sector using
emerging demand from stores within 60 minutes. vector autoregressive and vector error correction
We demonstrate feasibility of IoT system including models combined with the concept of Granger
image processing function and production causality. Based on the results the relationship is
schedule change decision making. Due to the bilateral, and the supporting explanation is quite
8
clear: high performing companies can afford costs and optimal stock policies in small and
holding higher inventories in order to run their large systems. The result was existing methods for
processes properly and provide higher service no backordering may lead to poor solutions in
levels to customers, all this resulting in higher firm case of partial backordering. The conclusion was
performance. The audience asked for other con- that partial backordering can be an effective
trol variables, such as time lag, but there is no means for rental systems to reduce lost demands.
data available. Industry effects will be consid-
The questions and comments for this presentation
ered in the future.
were as follows.
The third presentation is a work of three contribut- 1. Are Inventory costs for support depot and
ing researchers: Chiara Ronzoni, Andrea Grassi local shop assumed same?
and Andrea Ferrara (Universit degli studi di 2. Is one depot and one shop model excessive
Modena e Reggio Emilia, Italy). The aim of the simplification?
research is to address the optimal management
3. Demand function was given as normal distri-
of infrequent demand spare parts in the automo-
bution. The difficulty of optimal inventory poli-
tive industry, focusing on the distributor in the
cy will be in demand uncertainty.
supply network. The objective of the proposed
model is to allocate a set of products to a player 4. Usually local shops are not one. This study can
of the collaborative network minimizing the total be extended to multi local shop model.
costs equal for each player. The model is an ex-
tension of a previous one built by the authors. This
new one takes products obsolescence and new AUGUST 25, 2016 IM 6
product introduction into account. The model
by Erica Pastore
assumes the optimal management policy for the
products known. The work takes into account the
The sixth session on Inventory Management in-
variations in the mix of products.
cluded three presentations.
All in all, I feel unlucky to miss the session, the The first presenter, Erica Pastore from Politecnico
presentations seemed to be very interesting. I am di Torino (Italia), discussed an empirical analysis
very grateful to the speakers to provide me in- on the demand variability propagation (i.e., the
formation about their presentations for the pur- Bullwhip Effect) along a real three-echelon Euro-
pose of this report. pean spare parts Supply chain. Her main contri-
bution is the empirical prove that bullwhip is pre-
sent in the considered supply chain in different
AUGUST 25, 2016 IM 5 levels of aggregation of demand data. Also, sta-
tistical analyses on the causes show that retailers,
by Hisao Fujimoto when they are given incentives to forward-buy,
they tend to forward-buy fast moving products.
There were two presentations scheduled in this
session. However, second presenter has not at-
tended. So, we had only one presentation.
The presentation was titled Optimizing Stock

Levels for Rental Systems with a Support Depot
and Partial Backordering by Geriach van der
Heide.
This study treats the partial back ordering system

in the case of one support warehouse (depot)
and one local warehouse (shop) for estimating
9
The second presentation was held by Peter ers. For a selected set of customers, scheduled to
Chemweno, from the Centre for Industrial Man- be delivered in certain periods, a price discount
agement, Heverlee Leuven (Belgium) and dealt is offered if they agree to accept late, but guar-
with an adaptive risk assessment methodology anteed, deliveries. If a customer declines such an
for maintenance decision making. His model in- offer, it faces the possibility of a backorder. The
tegrates different frameworks, such as a Monte problem is modeled as a net present value Mar-
Carlo simulation to identify the equipment fail- kov decision process and is solved via a back-
ures, data mining to support the root cause anal- ward induction algorithm, exploiting the pri-
ysis and a decision scheme to select an appro- mal/dual properties of the model. A set of numer-
priate maintenance strategy for the recurrent ical experiments illustrates the properties of the
failures. optimal policies under a variety of operating
conditions. Finally, it is discussed as to how the
The third presenter, Hiroaki Ishii, who came from producer can induce the customers to reveal
the Kwansei Gakuin University, Sanda (Japan), their true due dates and demonstrate that cus-
tackled an inventory problem with processed tomers that wish to cheat cannot derive any
and non-processed products of same perishable benefits.
goods. His model sets optimal order quantities for
inventory with two different types of products: The second paper, titled Jointly optimal buyer-
one type is non-processed and has a single peri- supplier inventory decisions via third party coor-
od life time, while the other is a processed one dination (co-authored by Kurt Masten and Avijit
and has two period life time. Banerjee) was presented by Avijit Banerjee. This
paper points out the practical difficulties in im-
plementing jointly optimal integrated inventory
AUGUST 25, 2016 IM 7 replenishment decisions for a buyer-supplier pair,
where, of necessity, the two parties must openly
by Avijit Banerjee share all their relevant operating information.
More often than not, the parties involved are
The first paper scheduled in this session, titled
reluctant to share their respective proprietary
Backorders under advance demand synchro-
information. To overcome such reluctance, the
nized by price discount (co-authored by
role of a third party coordinator (3PC) is intro-
Muzaffer Alim and Patrick Beullens), was present-
duced in this study. This 3PC serves as a repertory
ed by Muzaffer Alim. This study focuses on a sin-
of all relevant information and acts as a facilita-
gle item, periodic review system, where custom-
tor for designing jointly optimal buyer-supplier
ers order the product ahead of time, to be deliv-
contracts, which are advantageous for the entire
ered by specified due dates. In order to amelio-
supply chain, compared to individual optimiza-
rate potential backorder situations, a demand
tion,. The joint optimization model incorporates
postponement mechanism is proposed, which
the risks borne by both the parties in making their
may benefit both the producer and the custom-
respective contractual commitments, which
necessarily lead to a closed end finite size con-
tract. A numerical example, together with exten-
sive sensitivity analyses, demonstrates the con-
cepts developed and the efficacy of this novel
approach.
The next paper, Dual channel inventory control

with cross-channel returns and transshipments
(co-authored by Arjan Dijkstra, Gerlach van der
Heide and Kees Jan Roodbergen) was presented
by Arjan Dijkstra. This work explores the possibili-
10
ties of transshipments in a dual channel supply spectively, a concave fractional programming

chain, that sells a single product over a specified procedure and a parameterized fractional pro-
selling season, consisting of multiple periods. gramming approach, are proposed. The proper-
Transshipments between the two channels are ties of the optimal policies are studied through a
allowed to occur, if necessary, on a periodic ba- set of numerical experiments.
sis during the selling season. An item sold through
a certain channel may be returned to either of
the two channels with known probability, the to- AUGUST 26, 2016 IM 8
tal returns being dependent on the products
market demand. The relevant costs considered A Linear Programming Framework for Control
are those pertaining to holding inventory, ship- of Tandem Production Lines by
ping and unsold stock at the end of the selling Ozge Buyukdagli - Murat Fadiloglu
season. A Markov decision process based model
is formulated for deriving the optimal values of
the initial stock levels at the two channels, as well
as optimal transshipments during the selling sea-
son. In order to avoid the computational burden
imposed by the Markov decision model, a lateral
transshipment heuristic is formulated. The perfor-
mance of this heuristic is compared to the opti-
mal policy, as well as other heuristic procedures
suggested in the existing literature. It is shown that
the suggested heuristic performs favorably com-
pared to the optimal policy, as well as the other
existing heuristic procedures. The analysis high-
lights the benefits of allowing cross-channel re-
In this paper the problem of production con-
turns in making transshipment decisions.
trol and stock rationing in a make-to-stock
The final paper of the session, Pricing and inven- production system with several customer
tory control for make-to-order service systems classes is considered that generate independ-
(co-authored by Ata Jalili Marand, Hongyan Li ent Poisson demands. The system consists of a
and Anders Thorstenson) was presented by Ata tandem production line with a finished goods
Jalili Marand. This study examines the issue of buffer at the end. At decision epochs, in con-
pricing, in conjunction with inventory control de- junction with the stock allocation decision for
cisions in a make-to-order service providing sys- the finished goods, the control specifies
tem. In order to respond to a customer request whether to continue or stop production at
for a specific service, it may often require some each workstation. The author modelled the
time to assemble or process the necessary parts, system as a Markov Decision Process (MDP)
materials, etc. in providing such a service. The and then proposed an equivalent linear pro-
existing literature has paid little attention to this gramming formulation.
phenomenon. In this paper, a simple queueing
The policy structure is further imposed by con-
process with Poisson arrivals and exponential ser-
straining the solution space via an innovative
vice times is employed to model the service sys-
introduction of a set of integer variables. Other
tem. With the service parts and other such items
than its theoretical attractiveness, the proposed
controlled by continuous review (Q, r) policies,
approach, which can be implemented on state-
while demand is modeled as a function of the
of-art optimization solvers, brings in considerable
price of the service, it is shown that the objective
savings in terms of computation time. Further-
function is pseudo-convex in structure. Conse-
more, the approach has the flexibility to be trans-
quently, two solution algorithms, based on, re-
ferred to many dynamic control problems.
11
Results of numerical experiments that cover both AUGUST 26, 2016 IM 9

optimal control algorithm and performance
evaluation models were presented in the study. by Gerard Reiner
The authors believe that there is a need for these
kinds of studies to understand the behavior of During the IM9 session on August 26 morning
production/inventory systems and to obtain val- three presentations took place.
uable insights about the dynamics involved. In
First, Steffen Klosterhalfen from the BASF SE, Lud-
addition to the characterization of the optimal
wigshafen in Germany presented a Control of a
policies, an efficient approach was proposed to
Continuous Production Inventory System with
optimize the parameters of the production con-
Production Quantity Restrictions. He conducted
trol mechanisms, that are very applicable in to-
this research together with his colleagues Falk
days manufacturing environments.
Holzbauer, from the Daimler AG, Mannheim, in
Germany and Moritz Fleischmann from the Uni-
"Thinner, lighter, faster ..." and the way forward for
versity of Mannheim, Mannheim, in Germany. The
time series extrapolation methods
author presented for a chemical industry produc-
by Konstantinos Nikolopoulos Assimakopoulos
tion system a control rule as well as an algorithm
Spiliotis Vassilios Assimakopoulos
for the control parameter determination that
Research in time series extrapolations methods is helps the planner to decide on the production
quite fragmented. Improvement in forecasting quantity in order to achieve a predefined service
performance is very detrimental and most bene- level target. Based on a numerical study the au-
fits in the process come from either better infor- thors demonstrated that the solution algorithm
mation sharing or management judgment. performs very well on a large set of parameters.
During the discussion at the end of the presenta-
Any new approach has to compete and beat tion the constraints and limitations were of par-
the two best established benchmarks in the field: ticular interest.
Damped Trend Exponential smoothing and the
Theta method dating from 1985 and 2000 respec-
tively. In this research paper it is discussed how a
new extrapolation should be developed via us-
ing elements and paradigms from the most suc-
cessful ones and we do propose such a new
method. The decomposition approach used in
the Theta methods is employed but instead of
using a linear trend we allow from more flexibility
and use a set of non-linear trends.
Furthermore instead of starting from the original

data smoothing first is appliedand thus there is a
much better starting point from the extrapola-
tion, something that the Naive method has
proved in many empirical investigations how im- Second, Gerald Reiner from the Universitaet Kla-
portant it is. The new proposition is evaluated in genfurt in Austria presented an Integrated In-
the M3 competition data with very promising sta- ventory and Capacity Management for Cabin-
tistical accuracy results. Bottom-line, for any new Based Transport Systems. He conducted this
method to compete with the existing bench- research work together with his colleagues Chris-
marks it has to be better at everything. tian Bettstetter, Pasquale Grippa, and Bernhard
Oberegger from the Universitaet Klagenfurt,
12
Thomas Schlechter from Skidata as well as An- Third, Arya Sevgen from the Izmir University of
dreas Kercek and Evsen Yanmaz for the Lakeside Economics in Turkey presented A Continuous-
Labs in Austria. The main contribution of the pre- Review Inventory Model with Disruptions and Re-
sented research work was to develop an efficient order Point. She conducted this research study
and effective access management, which regu- together with her colleague Fatma Zeynep
lates the entry of passengers to a cabin based Sargut from the Izmir University of Economics in
transport system via gates under joint inventory Turkey. They analyzed a two-echelon continuous
(waiting areas) and capacity constraints. The review inventory problem of a retailer and a sup-
objective is to reduce waiting times and improve plier.
fairness and, therefore, enhance the attractive-
ness of the system to, consequently, optimize the In particular it is an extension of the economic
long-term revenue (under consideration of cus- order quantity model with lost sales and reorder
tomer satisfaction, etc.) for the system operator. point when both supplier and retailer are subject
The transport system is modelled as a queueing to random disruptions. This problem is modeled
network with several interconnected queues as- with a Markov chain. The author concluded that
sociated with waiting and boarding areas in the if the ratio of supplier disruption rate to recovery
stations. The main questions after the presenta- rate is greater than one, a positive reorder point
tion was dealing with modelling assumptions re- can be optimal. The main questions after the
lated to the service time as well as potential ap- presentation were dealing with some model as-
plications of the research results to manufactur- sumptions. In particular the impact of the as-
ing systems. sumption that in case of SC disruptions all inven-
tory will be destroyed.
Inventory Management and Modelling for Perishable Product
AUGUST 22, 2016 PEPIM 1 by an MILP approximation of the SP model. Nu-

merical experiments compare the effectiveness
Fixed reorder days for a perishable product in of the policies to determine the order quantities
retail by Karin G.J. Pauls-Worm and Eligius M.T. with respect to costs and reached service levels.
Hendrix The first policy determines the order quantity con-
sidering the total available inventory. This gives
The authors study retail order policies for a prac- lowest cost solutions, which are not always feasi-
tical situation where stores may be able to order ble. From the three other studied policies, the
every day, or on fixed days of the week. Focus is policy where a fraction of the total available in-
on a perishable product which has a remaining ventory is considered when determining the or-
shelf life on delivery at the store of three days der quantity, performs best. One can always find
and a weekly (seasonal) demand pattern which a fraction that gives feasible solutions.
is stationary over the weeks. The stores have a
target -service level, i.e. the inventory should On the Benefits of Emergency Orders in Perisha-
cover demand in more than % of the days. The ble Inventory Systems by Chaaben Kouki, M.Zied
customer demand may be First In First Out (FIFO) Babai, and Stefan Minner
or a Last In First Out (LIFO) FIFO combination. A
Stochastic Programming (SP) model is presented The paper considers inventory systems for perish-
of the situation in the retailer practice. Four dif- ables with two-supply modes. Replenishments
ferent policies to determine the order quantity arrive at either regular lead-times or emergency
are studied. The base is a YS order policy where ones, characterized by a shorter lead-time but
the reorder days Y are fixed and order-up-to lev- with a higher purchasing cost. These policies use
els S are used, with parameter values generated real-time supply information that allows to derive
13
the steady-state behavior of the inventory sys- AUGUST 23, 2016 PEPIM 2
tems and obtain the exact optimal expected
total cost. The value of dual-sourcing in the con- by Chaaben Kouki
text of perishable items with stochastic lifetime
was also investigated. Pricing and advertising in supply chain of
perishable products under asymmetric
Near optimal ordering of perishables by Rene information by T. Avinadav and T. Chernonog
Haijema and Stefan Minner
This presentation considers a two-echelon supply
The ordering of perishable products in practice is chain for perishable. The authors investigated
commonly based on base stock policies (BSP), three wholesale price contracts which possibility
which are traditionally developed for non- of advertising. The channelcoordinating
perishable products. In the last decade, a num- conditions under each of these contracts are
ber of alternative order policies have been pro- demonstrated. Although all three contracts can
posed ranging from stock level dependent poli- coordinate the channel, potential
cies, which neglect the ages of products in stock, implementation issues with each of them are
to more complicated stock age dependent pol- discussed and guidance the retailer who want to
icies, which takes the ages of the products in design a contract units is provided.
stock into account. Recently a modified base
stock policy BSPlow is constructed that performs Suggestions are made by audience to consider
very well by adding parameters that will either other type of contact such as two-level buyback
smoothing order quantities or allow skipping or- contract, and a buy-back contract with channel
der moments. In this paper this policy is com- rebates.
pared with several stock-age dependent poli-
Dynamic pricing and dynamic shelf life to reduce
cies, and it is investigated what near optimal pa-
(Perishable) food waste at retailers by M. E.
rameter values of BSPlow are.
Buisman, R. Haijema and J.M. Bloemhof-
For a large design of experiments, optimal re- Runwaard.
plenishment parameter values are obtained by
simulation based optimization. These results can This presentation focuses on inventory manage-
be used to study the structure of near optimal ment of perishable products with dynamic
policies and to set near optimal parameters. pricing and dynamic shelf life. The main
objective of this research is to find solutions to
Numerical results obtained by simulation show reduce waste due to products perishability. The
that the policy BSPlow is doing even better than dynamic shelf life can be assessed by new
most existing stock-age dependent policies. technology called Time temperature indicators
Based on numerical insights, new stock-age de- that able to provide in real time the remaining
pendent policies are constructed and compared shelf life of product. The authors built a simulation
by simulation for a wide range of problem set- module to evaluate waste and profit using both
tings. Besides the comparison of policies we ad- dynamic pricing and dynamic shelf life They
dress how near optimal parameters can be ap- show that manning perishables with dynamic
proximated. shelf life and dynamic pricing or both allows to
A stock level dependent policy may outperform reduce waste increase profit.
a stock age dependent policy. Adding a waste
Testing Robustness of deterministic models of
correcting term may improve the performance
optimal pricing and log sizing models with
of a policy, but requires additional, usually com-
deterioration and partial backlogging by M.
plicated, calculations to be executed at every
Ghoreishi, C. Larsen.
order moment. For a fast determination of near
optimal parameters values, approximations are
The authors conduct a simulation experiment to
needed.
test the robustness of deterministic inventory
14
models with deteriorating items. They showed The second presentation was entitled Distribu-
that earlier research are indeed robust under tion Network and Inventory Decisions of a Food
stochastic inputs. Next, they developed a policy Aggregator cum Restaurant, by Kavita Chetana
iteration algorithm to find of optimal lost size for Didugu, co-authored by Chetan Soman.
deteriorating items with dynamic pricing. Finally,
The authors consider the operation of an online
they investigate numerically how changes in
hyperlocal food aggregator that functions also
price can affect the ordering quantities.
as a quick service restaurant (QSR). The distribu-
tion network decisions of the QSR are examined
at a strategic level, considering the partnering
AUGUST 25, 2016 PEPIM 3
restaurants and customers. The chosen restau-
by Karin Pauls-Worm rants belong to the same region as the aggrega-
tor. The authors propose a network design for the
According to the schedule, four presentations system, where the restaurants operate under ser-
were planned. Before the start of the session vice area constraints. There is demand informa-
there was a change in the schedule due to the tion sharing between the aggregator and the
cancellation by one author, leaving three pres- restaurants. All restaurants including the aggre-
entations. Unfortunately, after the second pres- gator decide how much of each product to
entation it appeared that also a second author produce and how much to outsource to each
cancelled the presentation. other. A Mixed Integer Programming model is
formulated to represent the network design. A
The first presentation was entitled Optimization numerical study is conducted by varying the
of Production-Inventory with Pricing and Promo- number of restaurants, the demand functions,
tion effort for a SVMB-VMI System of perishable the capacities of the restaurants, the range of
Products under Small Batch Shipment Policy, by products to offer and the maximum service ra-
Zhixiang Chen co-authored with Bhaba Sarker. dius. The model answers questions about the vi-
ability of the described business model and sup-
The authors investigated the mutual impact of
ports the analysis of the impact of the constraints.
production and marketing decisions and the
possibility of integration of production and mar- After the second presentation the session was
keting decisions for perishable products. The closed.
study considers a manufacturer and multiple re-
tailers. A VMI strategy is applied to control the
inventory. Pricing, promotion efforts,
the replenishment policy of the re-
tailer and the production lot size of
the manufacturer are taken as deci-
sion variables in the model. In order to
reduce the inventory of the retailers,
in the replenishment policy of the re-
tailers the JIT philosophy is applied.
The manufacturer delivers small
batches to retailers, where the trans-
portation cost is included in the total
cost. The authors compared central-
ized and decentralized models by
modelling two promotion strategies
where promotion cost were paid by
the retailers or shared by the manu-
facturer and the retailer.
15
AUGUST 26, 2016 PEPIM 4 search on food waste at retailers. There results
come from a European research project, which
by Rene Haijema involves 27 retailer of varying size. For the top 50%
of most perishable products they collected data
This parallel session, was the fourth session on in- on order quantities, pack size, etc. Statistical
ventory management and modelling for perish- analysis on the causes of food waste reveals that
able products. Unfortunately, the last of the three what they call the fresh case cover is a good
presentations was cancelled. predictor of food waste. The results that they
have presented give valuable insights.
The first presentation by Xiangyu Hou concerned
the display, inventory and disposal decision at a
Chinese wholesale market. The key problem ad-
dressed is how many products to display at the
booth and how many to keep behind in a
cooled area. A hierarchical Markov decision
problem is formulated that is solved by value it-
eration for a number of realistic problem settings.
For practical purpose, heuristics are derived and
compared to the optimal policy. The heuristics
seems to perform well over all experiments.
The second presentation was delivered by Karel

van Donselaar. He reported on empirical re-
Mathematical Modelling
AUGUST 22, 2016 MM 1 burg and it is co-authored by K. Inderfurth from

the same university. In this presentation new
by Gudrun Kiesmller approaches were presented to compute the
policy parameters of an (R; s; S) policy in case of
random demand and yield and positive
Three papers were presented in the first session leadtime. It was shown that the application of
on mathematical models at the international existing approaches only leads to acceptable
symposium of inventory research. The first and results in specific situations. Therefore, three new
the second paper had in common that they are approaches are presented, which have in
devoted to inventory models where fixed common that they include the uncertainty of
ordering cost were included. However, in the first outstanding orders in the inventory position. In a
paper a single unreliable supplier was considered large numerical study it was shown that the new
while in the second presentation two difierent approaches outperform existing ones.
supply options were possible, a normal one and
an emergency supply. In the third paper, even The second paper was presented by S.G.
more supply options were possible, because Johansen from Aarhus University. He considered
multiple-sourcing strategies and transshipment an inventory model, where additional to normal
strategies were compared. orders, expensive emergency orders are possible
in order to avoid expensive stockouts. In contrast
The first paper was presented by G.P. Kiesmller to other studies in this field, he additionally allows
from the Otto-von-Guericke University Magde-
16
the leadtime of the emergency order to be ran- and willingness or ability to pay on time. The pub-
dom. The cost performance of several lishers objective is to maximize the net present
replenishment policies is compared, where fixed value of the cash inflows from the contract. The
order cost and linear holding and backorder cost uncertain variables in the model are the quantity
are considered. It is shown that a policy where sold that is reported by the wholesalers and the
the normal order has always size Q and the quantity of units that the wholesaler ultimately
emergency order is based on information about pays for. The wholesalers cost functions depend
the inventory position and the time interval until on the decisions made by the publishers through
the next normal order is delivered has in general its own newsvendor problem. The problem itself is
an excellent performance. It outperforms policies analytically intractable due to the complexity of
where only one supply option is possible. the model, so the author solved the problem
numerically. He is currently testing the model with
M. Firoozi presented the third paper, written by real-world data obtained from a Hungarian pub-
himself and M.Z. Babai, W. Klibi, and Y. Ducq from lisher.
Kedge Business School Talence. The focus of this
presentation was on sourcing strategies in a multi The second presentation in this session was titled
echelon network where non-stationary uncertain Incorporating Parcel Transportation Costs into
demand is assumed. Besides the possibility to get Lot Sizing Decisions by Matthew Drake. The au-
material from difierent suppliers (multiple sourcing thor began the presentation by describing the
strategy) also lateral transshipments are allowed structure of parcel transportation rates, which are
in this model. For a given network structure zone-based rates that apply to shipments weigh-
difierent strategies are investigated and for the ing between 1 and 150 pounds. A graph of the
numerical comparison the problems are model- rates with respect to the shipment weight suggest
led as MIPs. The non-stationary uncertain that they increase at a nonlinear rate. This non-
demand is represented by difierent scenarios. linear rate of change does not have a functional
form, which makes it difficult to incorporate these
In a large numerical study the value of more costs into an organizations lot sizing decisions.
exibility in material sourcing could be illustrated. The research objective was to examine the effi-
cacy of a linear approximation to these parcel
rates. The author fit linear models to each of the
2016 FedEx Ground zone rates and developed a
AUGUST 22, 2016 MM 2
modified lot sizing model incorporating this linear
by Matthew Drake cost. He then conducted a numerical experi-
ment using nine different scenarios over the sev-
The first presentation in this session was titled In- en zones to compare the total cost of the lot size
vestigation of Initial Inventory Levels in a Supply obtained via the linear approximation to the op-
Chain by an Embedded Newsboy Problem by timal lot size with respect to the actual parcel
Imre Dobos. The author described the problem in rates. In most cases the optimality gap was 2% or
the context of the Hungarian book publishing less, suggesting that the linear approximation to
market and a small-to-medium-sized enterprise the parcel rates often performs quite well. The
(SME) publisher. This industry operates on a pro- optimality gap was only large in cases where the
ject basis with a high degree of fixed costs. Pub- optimal lot size lies in the drastically nonlinear por-
lishers and wholesalers (buyers) typically enter tion of the curve.
into consignment-type contracts with a fixed re-
tail price. In this scenario the wholesalers are in a The final presentation in this session was titled In-
powerful position compared to the publisher, ventory Decision and Pricing under Price-Based
and they differ in many attributes including their and Stockout-Based Substitution by Sandra
expected demand, margins, payment deadlines, Transchel. The author considers a firm that stocks
willingness to report sales timely and accurately, two products distinguished by quality attributes
17
such as regular versus organic. Consumers for the ties. The model is solved by using conic-quadratic
products exhibit assortment (price)-based substi- programming.
tution and stockout-based substitution prefer-
ences. Customer choice in this scenario is mod-
eled as a discrete choice. The firm must estimate AUGUST 23, 2016 MM 4
the customers quality valuation random variable
before making its decision. For a given low- A Variational Approach to the General Dynamic
quality product price, the stockout-based substi- Lotsizing Problem in Continuous Time by Robert
tution rates are endogenous. The author dis- W. Grubbstrm
cussed the solution procedure and then exam-
The Dynamic Lotsizing Problem, originally
ined the effects of changes in various problem
formulated by Harvey M. Wagner and Thomson
parameters, especially the elasticity of demand
M. Whitin (1958), considers determining the opti-
substitution.
mal size of batch quantities over time. This is a
generalisation of the Economic Order Quantity
Problem addressed by Ford Whitman Harris
AUGUST 22, 2016 MM 3 (1913).
by Stefan Minner From the original single-product formulation with

requirements as discrete time events, assuming
The session Mathematical Models 3 included no shortages permitted and instantaneous re-
two papers. Lena Silbermayr presented Invento- plenishments, through the years developments
ry pooling with environmental constraints using have been made relaxing several restrictions to
copulas, co-authored by Werner Jammernegg include backordering opportunities, allowing a
and Peter Kischka. The advantages of inventory finite production rate, multiple-product depend-
pooling in a newsvendor-type system are investi- ant demand in assembly systems, and several
gated under the additional consideration of re- other lines of extension (Andriolo, et al. 2014,
sulting emissions from transportation. Different Grubbstrm 2014A, 2014B, Grubbstrm et al.
forms of inventory pooling including transship- 2010).
ments are considered and compared to the de-
centralized no-pooling case. The additional Simultaneously the objective function has been
complexities that result from dependent de- generalised from the original Average Cost to the
mands are dealt with by using copulas. Net Present Value (NPV), from Hadley (1964) and
onwards.
The second paper on Network design with safe-
ty stock placement presented by Matias Schus- In the current paper, the author continues these
ter and co-authored by Stefan Minner and Jean- considerations treating the single-product case
Sbastien Tancrez models an integrated facility when demand varies over time continuously
location and inventory placement problem in a and/or in discrete impulses, applying the Calcu-
multi-echelon distribution-retailer network. Facili- lus of Variations and the NPV principle to formu-
ties for distribution centers serving retailers that late and solve the problem.
face stochastic customer demands are deter-
Planning for shortages? Net present value
mined in order to minimize operations, transpor-
analysis of models with backorder costs by
tation and inventory costs. The trade-offs ana-
Patrick Beullens and Yousef Ghiami
lyzed are the different opportunities to pool in-
ventories and to provide different service re- Should a firm plan its inventory replenishments
sponse times by planning different amounts of such that shortages purposefully occur? Can
inventories centrally or decentrally and by choos- such answers be found from deterministic
ing the number and location of distribution facili- average cost inventory models? It is well
recognised in the literature that unit costs of
18
backorders or lost sales are difficult to measure Interactions between flow of human resources in
and that this is a serious draw-back for getting functional regions of supply chain nodes and flow
real insight from such models. Since Grubbstrm of inventoris in processes of global supply chains
(1980) we know that answers to such questions by David Bogataj, Samo Drobne, Marija Bogataj
can be found by developing Net Present Value
models based on cash-flow functions. While this In the paper we shall demonstrate that the com-
technique has been used to gain insight into the bination of MRP Theory which enable to study
nature of holding costs, hardly any work so far flow of goods, and gravity models of labour
has been done on using this technique to commuting embedded in production functions
develop insight into the cost of backorders and of activity cells in the supply networks enable the
lost sales. mathematical formalization of Castells concept
known as reconceptualization of new forms of
spatial arrangements under the
new technological paradigm,
based on Ruggies work on Prob-
lematizing Modernity in Interna-
tional Relations and Arrighis fur-
ther study of interactions between
spatial entities and non-territorial
global economic space. Here the
idea of the functional region as a
unit in the space-of-places is con-
sidered. The nodes of a supply
chain are assumed to be in the
centers of functional regions. The
factors which influence availability
and costs of human resources are
studied on the bases of Lowry-like
model and the results are embedded in the pro-
A number of very plausible payment structures duction functions associated with nodes of a
are considered in backorder situations and are supply chain. The inventories in the process are
used to develop a series of NPV models which described as the items in the flow of MRP model.
can be compared with a series of classic models The aim is to construct a model which enables
with linear and fixed backorder costs, and with regional authorities and managers of global sup-
full and partial backordering. It is found that the ply chains to analyse and forecast the impact of
classic average cost EOQ models with shortages improvement of accessibility of nodes to human
are in general much more robust than the classic resources in functional regions on supply chains.
average cost EPQ models with shortages. The
specification of backorder and shortage costs in Here the net present value (NPV) of inventories in
EOQ models is intuitive and can be based on process in a global supply chain and human re-
only exogenous financial information. In EPQ sources in functional areas in the nodes of supply
models, the unit backorder and shortage costs chains are the criterion function. Therefore, the
are dependent on other model parameters such paper shows that studies of interactions between
as the annual demand rate, production rate, the supply chain flows (flow of items, information
and backorder rate. The average cost EPQ and financial flows) and human resources as
models need extra terms in their cost function in commuting flows in a functional region could
order to account for all opportunity costs and support a more sustainable economic growth.
rewards that depend on the decision variables.
19
given fixed order cost, and in the periods be-

AUGUST 23, 2016 MM 5 tween there are intraperiods, where there is a
higher fixed order cost. It was shown how to ex-
by Christian Larsen tend the concept of K-convexity in order to
make proofs for the structure of the optimal poli-
The session was attended by approximately 25 cy. It is of type (sj,Sj) for the intraperiods (and j is
people at each of the 2 presentations. The last an index for the intraperiod) while the optimal
presentation was cancelled. policy can be more complex for the regular or-
der periods. The theoretical findings were illus-
A heuristic for computing a new value-based
trated with a comprehensive numerical study.
stationary policy for the stochastic joint replen-
ishment problem by Christian Larsen
The presentation concerned the multi-item variant

of the stochastic joint replenishment problem. The
point of departure for the presentation was that by Yigal Gerchak
the Q(s,S) policy (considered in general to be
good) might be hard to interpret and furthermore Yigal Gerchak asked whether in a decentralized
might give bad results when featuring a set of supply chain the wholesale price is always
items with much heterogeneity among them. A decreasing in initial inventory level at retailer. He
heuristic modification of a given Q(s,S) policy (or a provided an example with random demand
can-order policy) was proposed. Its derivation where such is not the case. That questions the
used insights from the policy iteration algorithm. A universality of the desirability of carrying
numerical study showed that the new algorithm "strategic inventory".
does perform better than the Q(s,S) policy for
some selected datasets with heterogeneity Parisa Bagheri Tookanlou (joint work with
among them. Furthermore a numerical study also Hartanto Wong) discussed optimal customization
showed that the new policy can improve given levels in vertically differentiated markets. The
optimal can-order policies. model developed provides insights into the
conditions under which a quality-based
segmentation strategy is preferable.
Optimal stochastic inventory control with period-
ically varying fixed order costs by Florian Taube, Tatyana Chernonog (joint work with Tal Avinadav
Stefan Minner and Tal Ben-Zvi) discussed pricing and sales effort
in a supply chain of virtual products. Situations
with perfect and imperfect information regarding
the retailer's choice mechanism are analyzed.

by Sinem zkan
There were two papers presented in the seventh

session of the Mathematical Models of Invento-
ries stream. The session had an audience of
around 20 people.
The first paper in this session, A New Production

The presentation concerned a single-item setting
Policy for M/M/s Make-to-Stock System and Its
where there are regular order periods, with a
20
Renewal Analysis, was presented by Sinem AUGUST 25, 2016 MM 8

zkan (co-authored with nder Bulut). zkan
and Bulut study on a production setting of a sin- by Imre Dobos
gle-item, make-to-stock production system with
parallel production channels, fixed start-up cost The session has three presentations which have
and lost sales. zkan characterized the structure investigated mathematical issues of inventory
of the optimal production policy via numerical management. The papers have examined the
study and argued that the characteristic of op- classical problems of inventory control, such as
timal production policy is highly dynamic and not deteriorating items and optimal deliveries.
possess a systematic behaviour for all practical
purposes. zkan introduced their proposed pro- Abubakar Musa, Abdullahi Munirah, Bashir M.
duction policy which is well-performing and has Yakasai: Inventory Policies of Delayed Deteriorat-
well-defined structure. Their main contribution is ing Items with Stock Dependent Decreasing De-
to calculate the average cost per unit time for mand and Complete Backlogging
M/M/s make-to-stock queues with fixed costs
Thea authors have investigated a classical dete-
and lost sales under the proposed policy. They
riorating inventory control model with continuous
conducted the renewal reward analysis for
time. The demand for items is may be constant or
M/M/s make-to-stock queue with fixed cost in lost
stock dependent. The payment for these items
sales environment. Future extensions of the study
may be instantaneous or could be delayed for a
can be batch arrivals and batch production,
certain period of time offered by the supplier. The
multi-item or general processing times.
objective of the presentation is to develop an
The second paper, titled Emergency Service Lo- inventory control model of delayed deteriorating
gistics: Network Design and Dynamic Dispatch- items with stock dependent decreasing demand
ing, was presented by Herman Blok (co-authored and complete backlogging.
with Joachim Arts and Geert-Jan van Houtum).
Imre Dobos, Andrea Gelei: Dynamization of the
The paper examined emergency service logistics
Trust Game: An Inventory Control Approach
networks, such as fire truck dispatching services
and spare parts management. In order to over- The authors design an extended dynamic trust
come such incidents (fires, accidents, break- game incorporating both concepts trust and
down MRI-scanner or failure of a part), resources trustworthiness. Trustworthiness in this interpreta-
such as fire trucks and/or fire fighters or spare tion is a stock that accumulates resulting from
parts are need to be stored and dispatched on non-opportunistic behavior in risky situations. High
short notice. Blok emphasized three issues: (i) inventory level of trustworthiness in relationship
Where should resources be stored, and how results in action based trust. To analyze the
many resources need to be available at each above phenomena they develop a repeated
location? (ii) How should resources be dis- two person game, and present the analytical
patched in response to an emergency? (iii) Can patterns of behaviors of the players in the model.
the performance of the system be improved by
proactive relocation of resources? Blok made Epaminodas Kyriakidis, Theodosis Dimitrakos,
some assumptions to answer these questions Constantinos Karamtsoukis: Optimal Delivery of
such as dispatching from nearest point, no relo- Two Similar Products to N Ordered Customers with
cation, only one resource at each location, Product Preferences
symmetrical resource placement. In this talk, Blok
focused on planning at the tactical layer and The authors study a single vehicle routing prob-
analyzed how many resources should be availa- lem, in which a vehicle starts its route from a de-
ble at each location. Extensions of the study can pot loaded with items of two similar but not iden-
be nonsymmetrical placement of resources or tical products, which named product 1 and
available of relocation. product 2. The vehicle must deliver the products
21
to N customers according to a predetermined were provided for flexibility in volume as well as

sequence. The vehicle has a finite capacity and
after servicing all customers it returns to the de-
pot. It is assumed that each customer becomes
either product 1 or product 2 with known proba-
bilities. The travel costs between customers and
depot are known. The objective is to find the op-
timal routing strategy that minimizes the total ex-
pected cost among all possible routing strate-
gies.

variety.
by Zsuzsanna Hauck
Zsuzsanna Hauck: Optimizing the speed of pro-
This session included three presentations. Despite
duction vs. the speed of repairing in JIT environ-
the fact that we had an early start on the last
ment
day of the conference, quite a few attended the
session. All three speakers were quite young, the Zsuzsanna Hauck (University of Pcs, Hungary)
first of them being a PhD student, the others hav- presented her idea of extending the model of
ing defended their doctoral thesis last year. Vrs and Rappai (2016) considering quality in-
spection and repairing in JIT environment. Her
Chuanwen Dong - Sandra Transchel - Kai Hoberg:
assumption was that the speed of production
An inventory control model for Modal Split
and the speed of repairing can be a decision
Transport: A Tailored Base-Surge approach
variable for a firm to reduce backlogs and over-
Chuanwen Dong et al considered transport poli- production. She drew up and explained all pos-
cies enabling a shift in cargo volumes from fast to sible cases and gave deterministic formulas on
slow transport modes. They examined a Modal the optimal level of lot size, production speed
Split Transport model and introduced a Tailored and repairing speed.
Base-Surge approach with the aim to minimize
transportation, inventory holding, and backlog-
ging costs. A solution tool and numerical results
AUGUST 26, 2016 MM 10
were provided, based on which a lot of com-
Simple Modeling Techniques for Base-Stock In-
ments followed.
ventory Systems with State Dependent Demand
Luca Zeppetella - Elisa Gebennini - Andrea Grassi Rates by Fredrik Olsson
- Bianca Rimini: On the inventory distribution
In this paper new modeling techniques for (S-1,S)
problem with customer flexibility
inventory systems (continuous review base-stock
inventory systems) with state dependent de-
Luca Zeppetella (University of Modena and Reg- mand rates are proposed. Examples of single-
gio Emilia, Italy) and his colleagues considered location (S-1,S) inventory systems where the de-
the problem of a company that produces cus- mand, experienced by the system, varies due to
tomized storage tools from plastics. The speaker the state of the system are, e.g., inventory mod-
referred to Univer (the factory we visited in els with partial backorders, inventory models with
Kecskemt) whose supplier could use the model lost sales, inventory models with perishable items,
in practice. Solutions in make-to-order context inventory models with emergency replenishments
etc. Models for such inventory systems are in in
general hard to solve due to the fact that the
22
Markov property is often lost, and the prevalent on-hand inventory depleted, i.e., the rationing
tool used in the literature for providing exact solu- level for class 1 is zero. Arriving class 2 demands
tions for such models is partial differential equa- are instantaneously satisfied if the on-hand inven-
tions. Instead of using partial differential equa- tory level is above the critical rationing level of
tions with rather complicated analysis of bounda- class 2, which is denoted as K. Otherwise, class 2
ry conditions, we suggest considerably simpler demands are backordered. Backorders are
techniques which are based on elementary the- cleared according to the priority clearing mech-
ory of queueing and renewal processes. anism. That is, when a replenishment order ar-
rives, it is first used to clear class 1 backorders
First, new intuitive modeling techniques are pro- then inventory level is increased up to K and at
vided for (S-1,S) inventory systems with state de- this level, the remaining replenishment quantity is
pendent demand rates, which are considerably used to clear class 2 backorders. The inventory
easier than existing ones. Instead of introducing level can be increased above K only after clear-
heavy mathematical machinery as PDEs with ing all class 2 backorders. It is shown that if such
rather complicated analysis of boundary condi- an inventory system is sampled at multiples of
tions, which tends to cloud simple features of the supply leadtime, the state of the system evolves
model, alternative approaches are suggested according to a Markov chain. A recursive pro-
which are based on probabilistic arguments cedure is provided to generate the transition
concerning renewal processes and queueing probabilities of this embedded chain. Although
theory. The approaches reveal that it is possible the embedded Markov chain has an infinite
to derive a model which can track the location state space, it is possible to obtain the steady-
of the items in the system for general leadtimes. state probabilities of interest with desired accu-
By applying our approach, it is also possible to racy by considering a truncated version of the
further generalize the model to incorporate more chain. The obtained probabilities are also the
complicated demand interarrival patterns com- steady-state probabilities of the original continu-
pared to the standard Poisson demand arrival ous-time system and permit the computation of
process. any long-run performance measure of interest.
An Embedded Markov Chain Approach for the Exact Analysis of Lost Sales Base Stock Inventory
Analysis of (Q, R, K) Replenishment and Ration- Systems with Compound Poisson Demand
ing Policy by nder Bulut
Chaaben Kouki, M.Zied Babai, Zied Jemai, and
Rationing is an inventory policy that allows priori- Stefan Minner
tization of demand classes. It enables the inven-
The base-stock policy under the lost sales as-
tory system to provide higher service levels for
sumption has been extensively studied in the in-
critical demand classes. Companies often wish to
ventory literature. However, the optimal policy
provide different service levels to different cus-
parameter has only been determined under re-
tomers in order to achieve higher operational
strictive demand distribution assumptions. In this
efficiency. The underlying motivation usually is
paper, an earlier research is extended by provid-
the difference in service level requirements or
ing a simple recursive formula to compute the
shortage costs for customer classes. A backorder-
steady-state probabilities of the number of out-
ing inventory system is considered that experi-
standing orders for general compound Poisson
ences demands from two customer classes ac-
demands and general lead times. This enables to
cording to two independent Poisson processes.
numerically compute the exact total cost of the
Inventory is replenished according to (order base-stock inventory system under complete and
quantity, reorder-point) (Q, R) policy. It is as- partial rejection policies. Through a numerical
sumed that replenishment orders arrive after a investigation, we show the outperformance of
deterministic lead-time. If there is inventory on- our proposed approach.
hand, arriving Class 1 demands are instantane-
ously satisfied. They are only backordered when
23
Forecasting for Inventories
AUGUST 23, 2016 FC 1 the other hand, are update in very period, and
are more suitable for estimating the risk of obso-
by John Boylan lescence.
The stream on Forecasting for Inventories got off Detailed examination of the relative forecast ac-
to a good start on Tuesday morning, with a very curacy of the TSB and SBA methods has shown
well-attended session. The first talk was by Dennis that SBA outperforms TSB in some situations. In this
Prak (co-researchers Ruud Teunter and Aris paper, a Modified SBA method is proposed,
Syntetos) and focussed on the calculation of whereby the SBA formula is used after a demand
safety stocks. Dennis pointed out that the classic occurrence and a rule, based on the inter-
formula for normally distributed demand assumes demand interval relative to its forecast, deter-
both the mean and standard deviation to be mines the application of the TSB approach at the
known. In practice, however, neither of these end of periods with no demand. A simulation
quantities is known and, therefore, they must be experiment was conducted to compare the new
forecasted. method with previous methods, based on linearly
decreasing demand and sudden obsolescence.
The usual approach to this problem is simply to Forecast error was measured using Mean Error
plug in the forecasts into the classic safety stock and Mean Square Error. The Modified SBA meth-
formula. However, multiplying an estimate of od was found to outperform its competitors
demand variance per period by the lead-time is based on the simulation results. Experimentation
a flawed approach, as it fails to take into ac- on empirical data, using two real-world datasets,
count the autocorrelation of forecast errors, continued to show good performance of the
which may arise even if demand is, itself, inde- Modified SBA method but the improvements over
pendently distributed. Dennis demonstrated the other methods were more marginal.
correct formula to allow for error autocorrelation.
He also showed that, because the variance is
also being estimated, there are situations where AUGUST 25, 2016 FC 2
a t-distribution is more appropriate than a normal
distribution. The significance of the error of the by Aris Syntetos
plug in approach for inventory control has been
evaluated by numerical experiments based on This was the second session of the Forecasting
an (r,Q) policy and Simple Moving Average and for Inventories stream with a total of two sched-
Single Exponential Smoothing forecasting meth- uled presentations. Both papers considered issues
ods. It was found that large discrepancies may related to intermittent demand inventory fore-
arise in achieved service levels, particularly for casting. In particular, both papers looked at non-
short demand histories.
The second talk was given by Zied Babai (co-

researchers Selmen Boubaker, Rim Kalai-Jemai
and Yves Dallery). This talk concentrated on par-
ametric methods to forecast intermittent de-
mand. It began with a comparison of the
Syntetos-Boylan Approximation (SBA) and the
Teunter-Syntetos-Babai (TSB) methods for intermit-
tent demand forecasting. SBA forecasts are up-
dated when demand occurs, and are appropri-
ate as replenishment forecasts. TSB forecasts, on
24
parametric forecasting approaches for intermit- The Session 3 of Forecasting for Inventories
tent demand items. includes three abstracts and the three authors
were in the session to present these abstracts. In
The first presentation was made by Marwa Hasni what follows, Im providing a brief summary on
(joint work with Zied Babai, Mohamed Salah each presentation along with the feedback from
Aguir and Zied Jemai) and it was titled An inves- the audience and my comments.
tigation of bootstrapping forecasting methods for
intermittent demands. The researchers conduct- The abstract is entitled Demand Forecasting with
ed a detailed investigation of two bootstrapping Time-Irregular Stock Measurements for Oil, Gas
approaches that have been proposed in the and Chemicals and it has been presented by
literature. The first uses a Markov chain to switch Rogier Emmen. This work deals with a forecast
between demand and no-demand periods and problem in the context of inventory routing for
the second samples separately demand intervals Oil, Gas and Chemicals. The author considers a
and demand sizes. Both have been claimed to planning window of a day and short-term
provide improvements over parametric ap- forecasts are needed in the day before.
proaches but nevertheless, previous results have
been generated under quite restrictive assump- The author proposes a method to improve the
tions. In this work a numerical study was con- forecast accuracy and inventory performance.
ducted under more realistic assumptions fol- Through an empirical study using a real time seri-
lowed by an extensive empirical investigation. es from the context under concern, the author
The results were found to be conditioned shows the performance of the proposed method
(amongst other things) to the demand histories in terms of lower forecast errors and higher servi-
considered; when long demand histories are ce levels (and lower stock-outs).
available, the approach based on separate
The author has made a synthetic presentation
bootstrapping of sizes and intervals performs
(relatively shorter than the other presentations).
best. When short demand histories are available,
The problem was well motivated from the
the Markov chain based bootstrapping ap-
practical perspective which reflects well the
proach leads to the best performance.
knowledge of the author (as a practitioner) for
Following that, John Boylan, in a joint work with the practical context. Interesting questions from
Zied Babai, presented their paper on Non- the audience has been asked, especially on (i)
parametric estimation for intermittent demand: how the proposed method could be linked to
theoretical analysis and simulation results. The the theoretical background, and (ii) how this
authors considered three different non- work could be further extended to make it a
parametric approaches for estimating intermit- proper research paper.
tent demand requirements: i) non-overlapping
The abstract is entitled Incorporating Macro-
blocks aggregation; ii) overlapping blocks ag-
economic Leading Indicators in Inventory Mana-
gregation; iii) bootstrapping. Analytical results
gement and it has been presented by Yves
were presented on the properties of the boot-
Sagaert. This work deals with the issue of
strapping method, comparing its performance to
improving the forecasting and inventory perfor-
the aggregation methods, followed by a simula-
mance at the SKU level when macro-economic
tion experiment that allowed sensitivity results to
indicators are used. An empirical investigation is
be generated on various important control pa-
conducted based on a dataset from a business
rameters. The results open up opportunities for
to business company in order to show the per-
looking at many important issues in intermittent
formance of the proposed method.
demand forecasting.
The presentation of the authors was clear and
AUGUST 26, 2016 FC 3 the contribution of the work was well explained.
The author answered well to the questions of the
by Zied Babai
25
audience that were mainly dealing with the mance of the company departments under
proposed mathematical model, knowing that this concerns has been shown.
model was not easy to understand. The
This work is of a high value from the practical
presentation shows that this work was in an
perspective since it shows the results of a big
advanced stage of research and could be
research project with an implementation part.
submitted to the IJPE special issue of the
From the theoretical perspective, the topic is of a
symposium.
high value in the area of demand forecasting of
The abstract is entitled Development of spare parts and maintenance management. The
Forecasting Tool that Determines the Capacity presentation was also clear (although relatively
Requirements for Aircraft Maintenance Center dense with many numerical results shown to the
and it has been presented by Adel Ghobbar. audience.
There was only one question from the audience

This work deals with the issue of forecasting when
about the follow up of this work within the
the demand is intermittent or lumpy. It is a part of
company. If this work is well linked to the
a bigger project conducted with the Fokker
research background and the relevant literature,
Company in the Netherlands. Demand
this paper could be a good paper to be
categorisation and forecasting solutions have
published in the IJPE special issue of the
been evaluated and their impact on the perfor-
symposium.
Estimating Demand Uncertainty
AUGUST 22, 2016 DEMU 1 M/M/1 while numerical results are presented for
the M/G/1 case.
by Chulung Lee
The second talk was by Emel Arikan, who con-
This session on forecasting for inventories, saw two sidered various demand Models for combined
very excellent talks, and a good turnout with pricing and ordering decisions. They compare
nearly 20 participants attending. the standard approach based on deterministic
functions and stochastic error term, with directly
The first talk was by Hussein Naseraldin presenting specified distribution functions for demand which
work on the use of globalised robust optimisation depend on price. In her talk she presented ana-
as a tool for coping with irregular uncertainty de- lytical results regarding conditions of optimality. In
scribed as rare surges in arrival rates. His work addition examples were presented based on
considers three modelling approaches for multi- numerical simulation using specific demand dis-
period capacity planning, a nominal, robust and tributions, and the findings compared to the
globalised robust approach (GRO) to the M/M/1 standard approach.
and M/G/1 with the main contribution being the
redefining of the problem to include uncertainty, The final speaker was Nikolaos Kourentzes who
specifically irregular uncertainty, and the exten- presented simulation results as well as conducted
sion of the general solution for the case of global an empirical evaluation based on real time series
robust optimisation. Here irregular uncertainty is data, evaluating temporal aggregation for im-
defined as rare surges in arrival rate or what proving forecasting accuracy and decision mak-
Hussein referred to as the unknown unknown ing. The main contribution concerned the com-
and the GRO is evaluated as an approach for parison of multiple versus optimal temporal ag-
dealing with both regular and irregular uncertain- gregation levels and its impact on forecasting
ty. Analytic solutions are provided for the case of quality. The authors show that Optimal selection
26
while being optimal, is not robust enough to the proposed correction on meeting the invento-
model uncertainty at the original and aggregate ry targets.
level, and while multiple aggregation is by con-
struction suboptimal, it appear robust to such Asymmetric prediction intervals using half mo-
uncertainty and consequently produces more ments of distributions by Ivan Svetunkov, Nikolaos
reliable forecasts. Kourentzes
Correct prediction intervals are vital for inventory

AUGUST 23, 2016 DEMU 2 control, as they reflect the risks of going out of
stock or storing too much inventory. The conven-
Estimating demand uncertainty over multi- tional method of parametric intervals construc-
period lead times by Patrick Saoud, Nikolaos tion is based on the assumption of symmetry of
Kourentzes and John Boylan the error distribution. This assumption often does
not hold in practice and the observed forecast
Many business context decisions require demand errors, and the implied empirical prediction inter-
forecasts over a lead time in order to conform to vals, are non-symmetric and naturally non-
production and delivery requirements. Estimating Gaussian. This is particularly relevant when
the variability of lead time demand is crucial in events with high impact happen, such as promo-
setting safety stocks to hedge against demand tions or other disruptions in the supply chain,
uncertainty. Standard inventory literature pre- where skewed error distributions are expected. In
scribes the multiplication of the variance of the this case the deviation from normality makes the
errors by the lead time and the inverse cumula- conventional method of calculating prediction
tive distribution function of the normal distribu- intervals problematic. Alternative more compli-
tion. This approach however overlooks two corre- cated methods have been proposed, which
lations that exist within the forecasted demand: again do not necessarily guarantee that the cor-
the first is a correlation that arises between multi- rect intervals are constructed. We propose a
ple-steps-ahead forecast errors, and the other simple method of constructing prediction inter-
between the lead time forecasts themselves. As vals based on the half moment of the error distri-
a result, the safety stock levels reflect poorly the bution. Using this method allows to have both
actual quantities required as they underestimate asymmetric intervals in cases of skewed distribu-
the actual variance of demand. Despite these tions and the conventional symmetric ones in
correlations being flagged up by some authors, other cases. We conduct an experiment on real
no steadfast solution has been offered to over- data and show the efficiency of the proposed
come this issue, as these correlations can mani- method, and discuss implications for practice.
fest themselves in various forms. In this work we
first investigate these correlations from a theoret- Combining forecasts across different selection
ical standpoint to illustrate the magnitude of un- criteria by Devon K. Barrow - Nikolaos Kourentzes
certainty that is not accounted for. Following Fotios Petropolous
that, we propose various empirical approxima-
Nearly 50 years of research has focused on fore-
tions to remedy this problem. Although for certain
cast combination versus selection. From the one
demand processes an analytical correction is
hand findings suggest that correct identification
feasible to calculate, we provide a solution that
of the best model for a given series may lead to
is independent of the demand process. This is
significant accuracy improvements in some cas-
very relevant for practice, since identifying the
es up to 20-30%. Selecting the best method is
true unknown process is not possible and the
challenging in two ways: identifying the best
norm is that the forecasts are produced by a rea-
method given sampling uncertainties; and se-
sonable approximate model. We conclude the
lecting the appropriate criteria. For example,
work by demonstrating empirically the effect of
selecting an optimal model based on Akaike
Information Criterion (AIC) and Bayesian Infor-
27
mation Criteria (BIC) can yield different results. way to construct pools of forecasts to be com-
These challenges suggest that there are potential bined by combining across different criteria of
benefits from combining forecasts of models in- optimal models. Our approach combines fore-
stead. While simple combinations across all casts selected by different selection methods
available methods may not always be an ideal including, Akaike, Bayesian, Hanan Quin infor-
strategy, combinations performed across a set of mation criteria and cross-validation. We bench-
suitable methods or using appropriate combina- mark this strategy against the combination from
tion weights usually outperform the selection of a a single model selection criterion and evaluate
single model. Prior research has focused mainly the forecasting performance of the two ap-
on combining forecasts of different models, pa- proaches under different experimental setups
rameters or fitting samples e.g. through boot- providing recommendations for practice.
strapping. In contrast here we propose a new
Maintenance and Service Logistics
AUGUST 22, 2016 SERL 1 management cost and the disposal cost supplies
the stock. Since it is possible to supply when the
Stock Sharing Method between Regional Depots replenishment is needed, it is unlikely to become
for Spare Parts of Construction Machinery by again excess stock at the replenished depot.
Yuhei Niwa, Satoshi Sasahara, Kentaro Tagichi And minimizing the sum of these KPIs improve the
cash flow. Because, the purchase cost has a di-
The stock sharing between regional depots is rect effect on the stock reduction and other KPIs
one of the overall optimization methods of the have a direct effect on the profitability. In addi-
stock management in the pyramided supply tion, our method can evaluate the KPIs with the
chain. Normally, the stock sharing provides some weight that agreed the business strategy such as
good effects that are the reduction of stock the giving priority to the stock reduction, the sales
quantity, management costs and so on. On the promotion and so on.
other hand it provides the bad effect that is the
increase of the transport cost for the depot-to- The method was evaluated by simulation with
depot transport in the same hierarchy. Since the- actual order data of Hitachi Construction Ma-
se good and bad effects cancel each other out, chinery Co., Ltd.(HCM) which supplies more than
in order to maximize the effect it is necessary to 400,000 kinds of spare parts for excavators, dump
decide the timing/volume/from-to of the trucks and so on in their global pyramided supply
transport. In this study, we aim to improve the chain. As a result, the purchase cost was re-
consolidated cash flow that we have defined as duced by 3.8%, the transport cost was increased
the effect of stock sharing. by 60%, the stock management cost was re-
duced by 3.2% and the disposal cost was re-
The stock sharing method was developed be- duced by 20% than the general operation. The
tween regional depots for every replenishment transport cost was increased, but the profitability
order at depots. In the general operation, if the was almost equivalent to the general operation.
regional depot wants to be replenished the So the consolidated cash flow improved by
stock, the central parts depot (CPD) always sup- 10.5%. And the order fill rate is almost equivalent
plies the stock. In this method other depots, in- to the general operation.
stead of CPD, which have excess stock and can
minimize the sum of the 4 cost KPIs which are the
purchase cost, the transport cost, the stock
28
Aircraft Repair Process Efficiency Improvements: much easier to supply a spare part when the ship
How to Increase Reliability in a Component Pool- is in the harbor than when it is at sea. There are
ing System by Adel A Ghobbar - Isabelle Tozzi some simplifying assumptions, but the work is very
relevant in practice, and the author proposed
The maintenance of commercial aircraft is a interesting directions for further research.
necessity to ensure safe and efficient operation.
In an industry where the demand for safety is I, Rob Basten presented work on spare parts sup-
high, it is important that the maintenance is per- ply for planned and unplanned maintenance.
formed thoroughly and according to regula- We assume that planned maintenance is
tions. A downside to this story is that the current planned far ahead in time, so there is perfect
maintenance process is too long, which results in advance demand information. The unplanned
a specific maintenance process for each failure. maintenance leads to a stochastic demand
The objective of this research is to increase the stream. This means that we stock spare parts for
performances of the current repair process, two demand streams that differ also in their priori-
when outsourcing the maintenance, and lower- ties: planned maintenance has lower priority
ing the premature failure rate. The first part of than unplanned maintenance, since delaying
this research is to improve the performances of the latter directly leads to downtime. We com-
the repair process. The manufacturer is not per- pare the optimal spare parts inventory control
forming the maintenance themselves. The main- policy when planned maintenance may or may
tenance is done by subcontractors all over the not be delayed.
world. Three different groups were identified;
Administration, Logistics, and Repair Shop. With Rommert Dekker presented work on spare parts
the help of a general linear model, the signifi- inventory control for an installed base that is un-
cance of these three factors was able to be der a service contract. The question is what to do
calculated. The significance for Logistics and if the service contract expires: In practice, the
Repair Shop was high, for Administration it was expiration is usually not anticipated so that there
less significant, but still important. are many spare parts on stock after the contract
has expired, which are not used anymore. It turns
The second part of this research was to identify out to be better to start decreasing the stock
the premature failure components and introduce level already some time before the contract ex-
solutions for these components to increase their piration. This insight is applied in practice.
reliability. For five of these components failure
models were used to identify the critical parts
AUGUST 23, 2016 SERL 3
within each component. The models used are
Failure Modes and Effects Analysis (FMEA), Event by Qianru Ge
Tree Analysis (ETA) and Cause-Consequence
Analysis (CCA). Five components were subjected Physical asset management decisions in
to these models. In the end, all these compo- infrastructure companies by Behzad Samii
nents had problems with their design.
The presenter provided a well structuralized
framework of physical asset management
AUGUST 23, 2016 SERL 2 decisions in infrastructure companies integrating
elements like maintenance management, life
by Rob Basten cycle analysis, decision making framework and
so on together. All the concepts and the
Sena Eruguz presented work on the integration of frameworks were clearly explained. A case study
a preventive maintenance decision and spare of smart meter platform decisions was shown as a
parts supply. A key interesting feature of the case study in the end which completed the
model is that the installed base is moving. For whole topic.
example, a ship is either sailing or in a harbor. It is
29
Optimal Reliability and Commonality in Sajjad Rahimi Gharoodi presented a model to

Component Design by Joni Driessen Joachim plan both the availability of spare parts and of
Arts Geert-Jan van Houtum technicians to perform the replacement of such
parts. Such a model is intrinsically multi-
The presenter built a optimization model on the dimensional. For small problems exact evaluation
capital goods OEMs perspective to minimize the of system performance is possible by using matrix
LCC of the capital goods by using common plat- geometric methods. To solve realistic problems
form under the consideration of the availability of several performance evaluation approximations
the system. The problem is a realistic and can be are shown to work well by comparing to the true
applied to solve OEMs dilemma. The performance. Several optimization heuristics are
contribution of this paper is that it integrated also proposed based on a greedy logic and de-
three perspective/variables together in the decomposition between the planning of service
sign phase of the capital goods: the technicians and spare parts. It appears here that
commonality level of the platform, the reliability a greedy heuristic outperforms other approach-
of the component and the number of spare es.
parts. However, the author only use one
parameter the reliability impact to describe the Hasan Huseyin considers the integrated planning
cost difference between component produced of different repairable spare parts that require
by common platform and dedicated platform different skills for repair. The decision variables are
which is too ideal for a practical problem. the amount of spare part inventory, the number
of repair technicians and their skill-sets. Evalua-
Reliability Optimization for Multi-component tion of this problem is difficult and requires simula-
System in the Design Phase by Qianru Ge tion. Optimization is equally untractable so the
Zumbul Atan Geer-Jan van Houtum Ivo Adan authors propose a genetic algorithm for this pur-
pose.
The problem the present want to solve is to find
the optimal reliability of each component of a Joachim Arts presented a model to trade-off the
serial advanced system in the design phase stocking of spare parts and complete assets to
under a performance-based contract so that the meet a certain fleet readiness at minimal cost.
LCC of the system is minimized. The costs element Fleet readiness is the probability that sufficient
of the LCC is design cost, production cost, repair assets are operational at a random point in time
cost and penalty cost. The trade-off is clearly to fulfil the function of the fleet. Performance
explained as low reliability yield to low design evaluation turns out to be non-separable so that
cost, low production cost and high penalty cost. new algorithms for performance evaluation and
The relationship between reliability and repair optimization are proposed and shown to work
cost depends on the value of other parameters. well. Managerial insight is that stocking of assets is
However, the realistic meaning of divide the not desirable compared to stocking spare parts.
whole life cycle into small subperiods and than
find the optimal division for the total threshold
downtime needs to be further elaborated.
AUGUST 23, 2016 SERL 5
AUGUST 23, 2016 SERL 4 By Lucio Zavanella
By Joachim Arts Three interesting papers were presented and

widely discussed at the Session on Maintenance
This 4th session in the maintenance and service and Service Logistics, Session SERL5.
logistics stream covered three talks about inte-
grating multiple planning aspects within spare The first contribution, presented by E. van
parts inventory planning. Wingerden, discussed the case of lateral and
30
emergency deliveries of spare parts in a two- tween opportunities and drawbacks determined
echelon network. Once introduced the rele- by consolidation, as in the case of components
vance of the problem (e.g., randomness of the storage and repair lead times.
failures and high downtime costs), attention was
paid to the structure of the network itself (central,
emergency and multiple local warehouses, AUGUST 23, 2016 SERL 6
which may be requested for spare parts in se-
quence and admitting lateral transhipments). By Petros Boutselis
The objective of the study is the. total cost mini-
SERL6 session that took place on the last day of
misation. Thanks to local and central procedures,
the 19th ISIR Symposium included three presenta-
nested into an overall one, results are found and
tions. The first and the third suggested innovative
the need for optimisation procedure is outlined,
approaches of drawing information from the lo-
given the complexity of the problem and the
gistics support system to produce forecasts on
request for fast computation.
spares demand. The second presented the work
The second presentation, introduced by L. Za- that has been done on the effects analysis of off-
vanella, dealt with a case study observed in the line communication to the promotion of use of a
management of critical spare parts of complex load-matching system in transportations.
devices, in a Supply Chain consisting of one ven-
Mrs Sha Zhu from Erasmus University presented
dor, one supplier and the final user of the devices
the work that she did with Dr Willem van
(e.g., aircrafts). The need for high-service levels,
Jaarsveld from Eindhoven University of Technolo-
pursued jointly with the minimisation of total
gy and Professor Rommert Dekker from Erasmus
costs, was introduced, proposing a novel ap-
University on using the information that can be
proach to the management of spare parts, de-
provided by the maintenance plan about the
fined as Reverse Consignment Stock policy. The
defective components arrivals in order to esti-
structure of the costs in the Supply Chain was
mate the spare parts demand distribution. Using
discussed, consequently comparing the per-
this approach a periodic review, lost sales inven-
formance of the proposed policy (based on an
tory model was developed. The resulting dynam-
alternative cost sharing between the actors of
ic inventory model was tested in real data case
the Supply Chain in case of device failure and
studies and showed a reduction of around 8% on
consequent overhaul) with the performance of
the total costs. In the discussion period which fol-
the traditional approach. The comparison was
lowed, Mrs Sha Zhu explained that the lead-time
based on a preliminary case and set of data.
considered in the model was from a single period
The conclusive study, presented by N. Knofius, and also that considering the demand trend in
introduced the actual topic of 3D printing tech- history when estimating the failure probability
nologies, in the perspective of logistical implica- would increase the cost reduction according to
tions and related impact. The study firstly ad- the case studies.
dressed the technology and the concept of
Associate Professor Norihito Seki from Hokkai-
consolidation, i.e. the process of product re-
Gakuen University presented his study in investi-
design determined by 3D technology. Once out-
gating the effects of off-line / human communi-
lined the flow of materials in the repair-stock
cation information sharing on the promotion of
chain, the model (based on cost minimisation)
the use of the load-matching system Japan Lo-
was discussed together with the solution proce-
cal Network System (JL). The research included
dure. Therefore, a set of experiments, supported
the establishment of a structured questionnaire
by sensitivity analyses, was presented so as to
survey on 663 member corporations of 51 JL
respond to the critical question: is consolidation
braches. The resulting data were modelled using
always effective?. The interesting conclusions
structural equation models and verified the ef-
allowed to ponder on the existing trade-off be-
fect of human communication on the use of the
31
information system. Following the presentation variables to be considered for inclusion in the BN
Associate Professor Norihito Seki provided further and the theoretical support for the use of the
details on the sizes of the groups which partici- records kept in log-books as a sampled dataset.
pate in the off-line communication. The BN is then built in an unsupervised manner by
feeding the dataset into a score based algo-
The third presentation of the session was given by rithm. The BN model was compared to Single Ex-
Mr Petros Boutselis from Defence and Security ponential Smoothing (SES) and to Syntetos-
school of Cranfield University. This on-going work Boylan Approximation (SBA) in the prediction
under the supervision of Dr Ken Mcnaught of accuracy of 135 out-of-sample possible future
Cranfield University suggests an alternative mod- cases. The comparison showed promising results
el for forecasting failures when the context fac- in favour of the BN. In the discussion period Mr
tors that are either directly or indirectly associat- Petros Boutselis explained that the indeed difficult
ed with the failures are about to change. Using a task of eliciting the needed BN variables from the
simulation of the operations and support of some right experts can be facilitated by the use of the
generic UAV systems, data on the factors and first part of the methodology in the levels / nodes
the resulting failures have been collected for a of the Logistic Support Organisation that is mod-
period of 19 months and used in order to build a elled and from the records kept in their log-
static Bayesian Network (BN). The work is based books.
on two parts: a methodology that identifies the
32
Sustainable Logistics and Supply Chain Management
AUGUST 22, 2016 SUSL 1 ing Recycled Materials (co-authors: Lena

Sibermayr and Werner Jammernegg). Recycling
by Peter Kelle as an alternative supply option can increase the
economic, environmental and social sustainabil-
Christoph Glock from Technical University Darm- ity of supply chains. Though, quantities resulting
stadt presented the first paper Integration of from the recycling process are still uncertain and
Energy Aspects into the Economic Lot Scheduling therefore, demand cannot be met by this sec-
Problem (co-authors: Fabian Beck and Konstan- ondary source alone and the primary mining
tin Biel). The work integrated the energy aspects source is still needed. A single period inventory
into the well-known economic lot scheduling model was developed considering uncertain
problem (ELSP) by considering the product- prices for recycled materials, uncertain recycled
dependent energy usage during startup and quantities and uncertain demand as well as their
shutdown phases. They derived the optimal cy- potential dependencies. They derived the opti-
clic production schedule for both the Common- mal order quantities for recycled and primary
Cycle-Approach and the Basic-Period-Approach. materials and provide managerial insights in the
They proved that considering energy cost in ELSP economic and environmental benefits of multiple
clearly affects the cyclic production schedule sourcing with recycling. The supply security im-
and significantly reduces the companys energy proves by using the secondary sourcing option
costs providing improved sustainability. and enhancing the companys image by being
more sustainable.
The second paper, Evaluation of Sustainability
Tradeoffs in Multimodal Freight Transportation Each one of the presentations was followed by
Planning, was presented by Peter Kelle, from questions and interesting discussions proving the
Louisiana State University, USA (co-authors: importance of the sustainability in supply chain
Mingzhu Jin, Helmut Schneider and Christoph management.
Claypool). Since freight transportation is a major
economic driver and also a major pollution
source, it is important to find an appropriate AUGUST 23, 2016 SUSL 2
tradeoff in freight transportation planning. In or-
der to improve freight flow efficiency in a sus- Fuzzy assessed locations for urban cogeneration
tainable way they developed a simulation tool to plant as a node in the city's circular econo-
study the freight flow over all three major surface my, based on MRP approach by Danijel Kovai,
modes and their connections and, in turn, to help Janez Usenik and Marija Bogataj
the public officials identify the best way to im-
prove freight transportation. In a case study they The future of the smart city concept lays in gen-
demonstrated how to apply the proposed per- eral optimization of the whole circular economies
formance measurement system to evaluate the of the cities, where reduction of waste in ap-
mode tradeoffs considering environmental goals proach to production and reuse play one of the
and other performance measures.in Louisiana crucial roles. One of the effective ways to reduce
intermodal network for freight management. The excess inventories of waste on cities landfills is
studys major focus was the performance and to build cogeneration plants. Waste is trans-
sustainability effects of mode change between formed back to heat/energy which is used to
highway and train transportation. supply the city.
Patricia Rogetzer, from Vienna University of Eco- City consumes produced goods, and waste is
nomics presented the third paper Sustainable transported to cogeneration plant at location Li
Sourcing of Strategic Raw Materials by Integrat- where it is transformed to heat using the burning
33
processes. Distributed quantity of heat QD differs Sustainable Supply Chain Collaboration with Out-
from the actually usable heat received in the city sourcing Pollutant-Reduction Service by Huiping
(QR) due to losses throughout the distribution line Ding, Hua Huang and Qilan Zhao
(Q = QD - QR). Loses, maintenance costs and
transportation costs are all expected to rise with Over the past decades, reducing carbon emis-
the distance l between cogeneration plant and sion and pollutant has become consensus
the city. On the other hand, increasing the dis- worldwide. However, this complex job needs col-
tance l of cogeneration plant from the city laborative efforts of many related aspects, such
would make location Li more competitive due to as governments, enterprises, customers, and the
lower price of the land and lower direct impact community, etc. (Sommerville, et al. 2010). Pollu-
of the pollution to urban society of the city. tant and carbon emission reductions are particu-
larly interconnected with business behavior of
A model of input-output matrices is used to supply chain. How supply chain members would
mathematically describe material flows in such a collaboratively reduce pollutant and carbon
system, and Laplace transforms are used for ex- emissions has attracted high attention worldwide.
act introduction of lead times. This allows to fi- With this consideration, this paper intends to de-
nancially evaluate the model using Net Present velop a business model in which the operational
Value (NPV) principles. Here negative contribu- service of pollutant-reduction including equip-
tions such as pollution have negative price, and ment facility as a whole package is outsourced.
consequently negative contributions to overall In the model process, the supplier that provides
NPV. It was shown how to choose optimal loca- this service will bear the total cost of equipment
tion of cogeneration plant with NPV maximization facility installation and operational process, in
and apprise how successful can be a fuzzy ap- return for its service the supplier earns revenue
proach in such reasoning from a portion of electricity payment collected
by the power plant which depends on the sales
Impact of coordination on costs and emissions quantity of electricity during service periods. The
in a two-echelon EOQ model by Yann research focuses on the investigation of the
Bouchery, Asma Ghaffari, Zied Jemai, Tarkan mechanism that sustains collaboration between
Tan coal power plant and operational service suppli-
er by determining an optimal outsource price.
Coordination in supply chains consists in aligning
The crucial issue is through negotiating the out-
the decisions made by several echelons to reach
source price to reach the goal of reducing pollu-
a globally optimal solution called the centralized
tant and carbon emissions and meanwhile satisfy
solution, and to share the benefits among the
two partners economic interests.
actors. This concept has been studied widely
from a costs optimization perspective but coor- The results show that there exists an equilibrium
dination is also proposed by practitioners and outsource price that satisfies the supply chain
academics as a solution to reduce carbon emis- partners and meanwhile coal power plants per-
sions. This presentation focuses on the impact of formance are in compliance with the environ-
coordinating inventory decisions by comparing ment constraint. The research findings have the
the costs and carbon emissions resulting from the following implications. First, considering environ-
decentralized and the centralized optimal solu- mental externalities the governments supervision
tions of a two-echelon serial economic order should have incentives to motivate the coopera-
quantity (EOQ) model. Their model accounts for tion between the supply chain partners. Second,
transportation and inventory related costs and economic scale of output and sales price dis-
emissions and vehicle capacities. The count of the generated electricity power are the
authorsacknowledge that inventory decisions primary factors that is essential to the service
affect the load factor of the vehicle (also re- suppliers breakeven affects outsource price ne-
ferred to as the filling rate), and investigate if co- gotiation, which reflects the effectiveness of the
ordination can help increasing the load factors.
34
improvement of supply chains environmental In the last presentation entitled Procurement

and business performances. and Coordination under Imperfect Quality and
Uncertain Demand in Reverse Mobile Phone Sup-
ply Chain, Amy Zeng of Worcester Polytechnic
AUGUST 23, 2016 SUSL 3 Institute presented a mathematical model that
helps to coordinate supply chains for used mo-
by C. Glock bile phones. The focus of the model was on ana-
lyzing the interplay between sales quantity dis-
The focus of session SUSL3 was on purchasing
counts and quality improvement investments for
patterns of small and medium-sized companies,
the case where the quality of used mobile
the management of time-slots for warehouse
phones is uncertain. The results of the model indi-
loading docks and the coordination of reverse
cate that the initial imperfect quality percentage
supply chains for mobile phones.
of mobile phones is crucial for the coordination
In the first presentation entitled Positioning the mechanism that should be selected to maximize
State of Purchasing in Small and Medium Sized the profit of the supply chain.
Enterprises (SMEs), Andrew Yao of California
State University presented the results of an empir-
ical investigation on the purchasing behavior of AUGUST 25, 2016 SUSL 4
small and medium-sized companies and used a
purchasing portfolio approach to analyze the by Danuta Kisperska-Moro
data. The results of the study indicate that the
majority of SMEs in the sample have a strong fo- Three papers were presented during that session:
cus on an effective cost and materials flow
First paper titled Competitive Advantage of Dairy
management, and that significant differences in
Enterprise Supply Chain: Perspective of Quality
the purchasing behavior of SMEs and large com-
and Safety presented by Huiping Ding was fo-
panies exist.
cused on the problems of quality and safety in
The second presentation entitled Time Slot dairy industry in the context of competition of
Management Systems for Warehouse Loading foreign companies in China. The authors ad-
Docks A Simulation-based Analysis of the Be- dressed the following research questions: what
havior of Warehouse Operators focused on the are the reasons of advancing quality and safety
impact of different behaviors of warehouse op- in dairy supply chains, what factors affect quality
erators displayed as a response towards missed and safety in dairy supply chains and how those
delivery time slots. Dominik Thiel of Technische factors impact the growth of competitive ad-
Universitt Darmstadt first classified different be- vantage of Chinese dairy supply chains. The
havioral patterns that had been identified in a main research objective was to identify interac-
review of the literature and during interviews of tive safety behavior of suppliers in dairy supply
shippers in Germany. Subsequently, he reported chains and relate them to corporate social re-
the results of a simulation model that tried to es- sponsibility. The Authors used structural equation
timate the cost impact of the different types of model based on analysis of data collected
behaviors identified in the first step. The results of through a questionnaire, expert interviews and
the study indicate that time window constraints running AMOS software. Created causal model
and the behaviors of warehouse operators have of structure of dairy supply chain allowed to con-
a significant influence on the total cost of road firm several hypotheses. As a result it was proved
freight transportation, which implies that the be- that quality and safety constitute the foundation
havior of warehouse operators should be consid- of competitive advantage in dairy supply chains.
ered when planning delivery time windows.
Second paper titled Inventory Policies for Energy
Storage Systems Optimization, presented by
35
Simone Zanoni, was based on application of in- cal Study of Sustainability in the Oil and Gas In-
ventory theory to energy storage systems (ESS) dustry. They exposed the requirements to under-
which seem to be the important technology of stand the drivers and barriers to sustainability in
the future. The objective function used for the the sector. Based on the questionnaire their re-
optimization was the minimum cost of energy sults indicate that there are several drivers of sus-
storing system. The costs related to inventory and tainability like energy conservation, resource de-
supply chain management were translated into pletion, but legal and regulatory pressure was not
the electric energy storing model. The main con- found as important factor. The main inhibitors of
clusion was that energy storing systems present sustainability practices were listed like inappro-
similarities with inventory systems. The model ap- priate facilities, unskilled human resources, insuffi-
plied during the course of research allowed to cient information and high cost of adoption.
find the optimal strategy of operation of the ESS.
Danuta Kisperska-Moron reported on research
The third paper of Gyngyi Vrsmarty addresses results based on a survey concerning the rules
the problem of green supplier selection. Litera- and standards of organizational governance in
ture review on criteria of evaluation has been different types of companies operating in supply
completed and it revealed that since 1990 chains. In her presentation of this research under
green criteria were added. Literature review the title: Corporate Governance as a Factor of
considered also the methodology of evaluation Building Social Responsibility Supply Chain Per-
and it allowed to classify it into the groups of pa- spective she presented the answers of Polish
pers basing on application of such methods and managers on two corporate social responsibility
also evaluation according to practical applica- questions:
tion with the cost being a major preferred factor.
The Authors indicated the challenges facing RQ1: How the type of a company and its role in
evaluation of suppliers, and mainly stable suppli- the supply chain is related to organizational gov-
ers base and costs consideration (data availabil- ernance for corporate social responsibility and
ity and defined processes). The aim of the paper
RQ2: How the size of a company is related to
was to provide a rank and the best supplier
organizational governance for corporate social
evaluation model which would be easy to han-
responsibility.
dle and provide quick results. Proposed model
included input indicators such as management The results of data analysis showed that there are
indicators (quality, price, lead time, etc.) and noticeable differences of strategic managerial
output indicators (environmental, reusability, CO2 attitudes towards governance solutions regard-
pollution, etc.). Linear programming was used for ing social responsibility depending on the type
the supplier evaluation model. Inventory related and size of a company. She explained that large
costs were included in the input indicators. companies and manufacturers seem to adopt
adequate strategic solutions more often. Prof.
All papers presented new interesting results and
Kisperska-Moron collected the list of recommen-
provoked the audience to widely discuss pre-
dations which can be very useful to other man-
sented problems.
agers in each specific area.
In their paper: Pricing Decisions and Coordination

AUGUST 25, 2016 SUSL 5 in Closed-loop Supply Chain with Patent Licens-
ing and Remanufacturing Competition, three au-
by Marija Bogataj
thors: Juhong Gao, Meng-meng Li and Jin-hui
Yahaya Y. Yusuf presented results of a study in oil Teng studied Close Loop Supply Chain (CLSC)
and gas industry achieved together with Jan systems. Based on Stackelberg game theory ap-
Hewitt, Tijjani Abubakar, Louise Mc Ardle and proach they developed pricing model in order to
Wendy Auchterlounie, under the title An Empiri- investigate the effects of patent licensing fees
36
and remanufacturing competition when de- ing problem in a green supply chain was can-
mand is uncertain on pricing decision and coor- celled by the author Rekik, Yacine.
dination mechanisms. They conclude that from
the perspective of CLSC management, manu- Thus, there were only two papers presented in
facturers should decide the rational patent li- this session including:
censing fee and maintain the appropriate level
1. Service capacity procurement of logistics
of remanufacturing competition in order to
service supply chain with demand updating
achieve an efficient operation and profit maximi-
loss-aversion behavior, authored by Liu,
zation goal.
Weihua.
Pattana Boonchoo and Suthep Nimsal studied 2. Sustainable decision model of flexible JIT
logistics integrations as presented in their paper: assembly procurement-production-inventory
The Logistics Integration of Small Scale Farmers in with recoverable items and emissions con-
Modern Supply Chain: The Role of Social Enter- straint, authored by Chen, Zhixiang.
prise and Sustainable Design in Thailand. The ob- The above two presentations discussed interest
jective of their study was to assess the impacts of issues with backgrounds from industrial practices.
logistics integration of small-scale farmers in
The first presentation intends to address express
modern supply chains, emphasizing the role of
delivery service capacity problem that logistics
social enterprises. As they reported, the findings
service providers face with, particularly when
indicate that there were many reasons for partic-
demand for express delivery of goods purchased
ipate in such integrated supply chain. Social en-
through internet is high.
terprise helps to reduce logistics including inven-
tory costs and increase their productivity. The second presentation intends to investigate
optimal decisions for using recoverable items
(used beer bottles) and reducing production
AUGUST 26, 2016 SUSL 6 emissions in the manufacturing process of brew-
ery industry.
by Huiping Ding
During the presentations, several questions from
There were three paper presentations scheduled audiences were interactively discussed in this
in this session, however, the paper presentation session.
with the title A multi-objective robust stochastic
This session was held on time and successfully.
model for a transshipment-enable inventory rout-
37
Session on the Best Student Paper Award finalists
by Sandra Transchel
On Thursday, August 25, 2016 the three finalists of capacity in which the system naturally exhibits
the first ISIR Best Student Paper Award (BSPA) load-dependent replenishment lead times.
were presenting their papers in the session. The
three finalist were Christians research was at the interface be-
tween Operations Management and Finance.
1. Hamed Jalali, Raisa Carmen, Inneke Van He investigated an inventory control model for
Nieuwenhuyse, Robert Boute (KU Leuven, commodities and raw materials that are bought
Belgium) at spot markets and that are characterized by a
Assortment and Pricing in Produc- high price volatility with unobservable states. To
tion/Inventory Systems model price uncertainty with unobservable
states, Christian used hidden Markov regime
2. Christian Mandl, Stefan Minner (Tech- switching technique, which is a common meth-
nical University of Munich, Germany) odology to deal with unobservable states in sto-
Commodity Spot Market Procurement chastic processes. He proposed an approach
under Hidden Markov Modulated Price where price observations at the spot market are
Regimes considered in order to update probabilistic price
regime information in a Bayesian way. Various
3. Danja Sonntag, Gudrun P. Kiesmller research questions were addressed in Christians
(University of Magdeburg, Germany) paper, for example: What are the effects of un-
The Influence of Quality Inspections on observable price regimes on the structure of the
the Optimal Safety Stock Level optimal inventory policy? How do price regimes
affect the monotonicity behavior of the inventory
All three papers demonstrated high quality re- policy parameters? What is the cost of price pro-
search, conceptual and methodological rigor, cess misspecification?
practical relevance, novelty, and potential im-
pact to their respective fields. Danjas research was related to the field of pro-
duction and inventory planning with random
Hamed investigated in his paper the impact of yield issues. She considered a multi-period make-
load-dependent lead times on the optimal as- to-stock serial production systems of a single
sortment and pricing decisions of a manufacturer product with in-between quality controls (inspec-
who offers an assortment of vertically differenti- tions) of the produced items. Assuming stochastic
ated products. i.e., products differ in their quality proportional yield, Danja showed first how to
valuation. Hameds research was at the interface compute the optimal safety stock for such a sys-
of Operations Management and Marketing and tem, second how to determine the optimal loca-
he studied a highly relevant and innovative prob- tion of one inspection, and third how to deter-
lem. Hamed specifically considered a make-to- mine the optimal number and the optimal posi-
stock manufacturer who needs to decide on the tion of quality inspections across the production
price, quality, and inventory levels of the different periods. Her major findings were that the safety
products offered, in view of maximizing his ex- stock level heavily depends on the shape of the
pected profit. He characterized structural proper- yield across the production periods, i.e., it highly
ties of the optimal decisions assuming infinite depends on whether yields are increasing or de-
production capacity and shows that for normally creasing across production. Danja furthermore
distributed demands, the optimal decisions have showed that marginal cost reduction of every
a simple structure. He compared these results to additional worthwhile quality inspection is rapidly
the more realistic scenario of finite production decreasing.
38
The winner of the Best Student Paper Award was

Danja Sonntag
The runner-ups were

Hamed Jalali
Christian Mandl
C o n g r a t u l a t i o n s !
39
Moments of
the banquet
evening
Attila Chikn Ou Tang Ruud Teunter, ISIR Presidents

welcomed the participants inauguration speech
40
Awardees of the Symposium

ISIR Fellows elected new Fellow
Stefan Minner
ISIR Service Award 2016 went to

Avijit Banerjee
and Hiroaki Ishii

who had to leave earlier.
41

ISIR Symposium 2016 Session Reports

Uploaded by

Copyright:

Available Formats

ISIR Symposium 2016 Session Reports

Uploaded by

Document Information

Original Description:

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

ISIR Symposium 2016 Session Reports

Uploaded by

Copyright:

Available Formats

19th International Symposium on Inventories

August 22-26, 2016 Hotel Hilton Budapest, Hungary

Session chairs report

AUGUST 22, 2016 distributions. Examples are used to illustrate the

Maximiliano Udenio, Kai Hoberg and Jan

In this presentation, Rommert Dekker presents the

Opening Plenary Session 2 nance decisions jointly with spare-parts inventory

The second paper presented, Joint condition-

The three presentations were most interesting,

AUGUST 22, 2016 EC1 bles for industry (services/manufacturing) and

Order Picking System is a warehouse manage-

Introduction of IoT system to real time produc- AUGUST 23, 2016 IM 4

The presentation was titled Optimizing Stock

This study treats the partial back ordering system

The next paper, Dual channel inventory control

ties of transshipments in a dual channel supply spectively, a concave fractional programming

Results of numerical experiments that cover both AUGUST 26, 2016 IM 9

Furthermore instead of starting from the original

Inventory Management and Modelling for Perishable Product

AUGUST 22, 2016 PEPIM 1 by an MILP approximation of the SP model. Nu-

The second presentation was delivered by Karel

AUGUST 22, 2016 MM 1 burg and it is co-authored by K. Inderfurth from

by Stefan Minner From the original single-product formulation with

given fixed order cost, and in the periods be-

The presentation concerned the multi-item variant

AUGUST 25, 2016 MM 7

There were two papers presented in the seventh

The first paper in this session, A New Production

Renewal Analysis, was presented by Sinem AUGUST 25, 2016 MM 8

to N customers according to a predetermined were provided for flexibility in volume as well as

AUGUST 26, 2016 MM 9

Forecasting for Inventories

The second talk was given by Zied Babai (co-

There was only one question from the audience

Estimating Demand Uncertainty

Correct prediction intervals are vital for inventory

Maintenance and Service Logistics

Optimal Reliability and Commonality in Sajjad Rahimi Gharoodi presented a model to

AUGUST 23, 2016 SERL 4 By Lucio Zavanella

By Joachim Arts Three interesting papers were presented and

Sustainable Logistics and Supply Chain Management

AUGUST 22, 2016 SUSL 1 ing Recycled Materials (co-authors: Lena

improvement of supply chains environmental In the last presentation entitled Procurement

In their paper: Pricing Decisions and Coordination

Session on the Best Student Paper Award finalists

The winner of the Best Student Paper Award was

The runner-ups were

Attila Chikn Ou Tang Ruud Teunter, ISIR Presidents

Awardees of the Symposium

ISIR Service Award 2016 went to

and Hiroaki Ishii

You might also like