combining make-to-order and make-to-stock inventory policies: an empirical application to a...
TRANSCRIPT
This article was downloaded by: [University of Southern Queensland]On: 09 October 2014, At: 11:54Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Production Planning & Control: The Management ofOperationsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/tppc20
Combining make-to-order and make-to-stock inventorypolicies: an empirical application to a manufacturingSMEMarco Perona a , Nicola Saccani a & Simone Zanoni aa Department of Mechanical and Industrial Engineering , University of Brescia , Via Branze38, 25123 Brescia, ItalyPublished online: 15 Sep 2009.
To cite this article: Marco Perona , Nicola Saccani & Simone Zanoni (2009) Combining make-to-order and make-to-stockinventory policies: an empirical application to a manufacturing SME, Production Planning & Control: The Management ofOperations, 20:7, 559-575, DOI: 10.1080/09537280903034271
To link to this article: http://dx.doi.org/10.1080/09537280903034271
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Production Planning & ControlVol. 20, No. 7, October 2009, 559–575
Combining make-to-order and make-to-stock inventory policies:
an empirical application to a manufacturing SME
Marco Perona, Nicola Saccani* and Simone Zanoni
Department of Mechanical and Industrial Engineering, University of Brescia, Via Branze 38, 25123 Brescia, Italy
(Final version received 17 April 2009)
This article focuses on decoupling point and inventory policy decisions in manufacturing companies supplyingproducts with different demand patterns and customisation levels. In such a context, adopting a pure make-to-order (MTO) approach may severely affect the response time for standard and regular products while, on theother hand, a pure make-to-stock (MTS) policy may result in excess inventory. To overcome this, companies tendto adopt hybrid and dynamic MTO–MTS policies, but decisions are often taken without the support of a rationalmodel. In this article we develop a rational model to support inventory management decisions in a MTO–MTScontext and bridge the gap between theory and practice. Starting from a real-life case study, we develop adecision-making approach that employs simple models, methods and tools, thus making it suitable for practicalimplementation in small and medium sized enterprises (SMEs). Different product characteristics are analysed inorder to develop a framework for choosing the most suitable decoupling point and replenishment policy (such aseconomic order quantity, EOQ) and for determining the parameters of the chosen policy (such as lot size).The simplicity of the procedure together with the positive results achieved in this first case study implementationsuggest that the new framework has the potential to improve the inventory policies adopted by SMEs in aMTO–MTS context and should be refined and developed through further case study research.
Keywords: make-to-order; make-to-stock; small and medium sized enterprises; implementation; case study
1. Introduction
Inventory planning in multi-product and multi-stagemanufacturing systems is challenging – the stage atwhich the decoupling (or order penetration) point(Hoekstra and Romme 1992, Olhager 2003) is posi-tioned has to be determined for each product. Oncedetermined, an inventory policy has to be chosen inorder to set the inventory level and reordering policy atthe chosen decoupling point. Moreover, each inventorypolicy comes with a set of parameters to be tuned: forinstance, an economic order quantity (EOQ) and areorder point (OP) have to be set in order to apply theEOQ–OP model (Mennel 1961, Silver et al. 1998). Thisis a multi-dimensional problem: differences in the maindecisional drivers (for instance, the cost structure orthe demand pattern) occur along the transformationprocess, across the product mix and over time.
As a consequence, one single policy is unlikelyto fit all these features: a hybrid policy combiningdifferent decoupling points, inventory models and/orparameters for different products could instead beadvisable (Soman et al. 2004, 2007). Yet, on this point,the approaches adopted by researchers and practi-tioners tend to diverge. The theoretical approaches
focus on joint multi-item and multi-stage cost optimi-
sation but are either too complex to be effectively usedin real cases or fail to consider all relevant constraints.Therefore, practitioners tend to base their planningdecisions on common sense and experience without anunderlying rational model (Federgruen and Katalan
1995). These flexible approaches allow the practitionerto consider all relevant parameters and constraintsbut they often lack the required consistency. Hence,there is a gap which very few have tried to bridge(Van Donk 2001).
This article aims at bridging the aforementionedgap between theoretical models and current industrial
practice. The majority of academic research on pro-duction planning and control (PPC) starts from theoryand then tries to put it into practice (see for instanceStevenson and Silva (2008)); yet theoretical methodsmight be at odds with reality. This may happen
because the required data are not available, the shop-floor managers are not skilled enough, some relevantconstraint has not been taken into consideration, orsimply the addressed problem is not commonly foundin reality (Perona and Miragliotta 2000). In this study,
instead, we start from the shop floor and attempt to
*Corresponding author. Email: [email protected]
ISSN 0953–7287 print/ISSN 1366–5871 online
� 2009 Taylor & Francis
DOI: 10.1080/09537280903034271
http://www.informaworld.com
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
determine the incremental improvements needed to
bridge the quantitative and rational decision-making
gap between the theoretical methods and everyday
practice.Given the above, the remainder of this article is
organised as follows: Section 2 provides a critical
assessment of the available literature in light of the
above; Section 3 presents the methodological approachthat has been followed in this study; Section 4
contains the description of the company where the
case study has been developed; Section 5 describes the
new planning method proposed in this article; and
Section 6 illustrates the comparative advantages it
achieves in the specific case considered. Finally,
Section 7 contains a critical discussion of the workundertaken and suggestions for future research.
2. Literature review
A major problem faced by managers in multi-stage
manufacturing contexts is the positioning of the
decoupling point for each product (Hoekstra and
Romme 1992). As highlighted by Olhager (2003),
different positions of the decoupling point define
different ways of fulfilling commercial demand, withan impact on such performances as customer service
level, manufacturing efficiency and working capital
tied up in inventory. In the literature review presented
in this section we consider papers which explicitly
address the decoupling point positioning problem in
hybrid make-to-order–make-to-stock (MTO–MTS)
production systems, while we do not consider thosefocusing only on production planning or scheduling
problems, such as Federgruen and Katalan (1999),
Tsubone et al. (2002), Wu et al. (2008). We categorise
the relevant papers according to the main approach
adopted to analyse the decoupling point positioning
problem as analytical models, frameworks, or case
studies, with some belonging tomore than one category.An overview of the literature is shown in Table 1.
Although in real life many multi-stage manufacturing
contexts adopt hybrid MTO–MTS inventory policies,
Soman et al. (2004) point out that literature addressing
these systems and their peculiar problems is still scarce,
while most research deals with pure MTS or MTO
systems. This is also our finding, demonstrated by the
fact that Table 1 encompasses only 11 contributions,scattered over 16 years. Table 1 highlights the main
contribution to theory and/or practice of each paper,
as well as the main limitations preventing the real-life
application of the solutions proposed or making it
difficult.
Six papers in Table 1 propose an analyticalmodel, pursuing a monetary objective function, suchas the minimisation of a cost function (Arreola-Risaand DeCroix 1998, Rajagopalan 2002, Gupta andBenjafaar 2004, Kaminsky and Kaya 2008) or themaximisation of the expected profit (Li 1992). Servicelevel is considered either in the objective function(through backorder cost) or as an external constraint(maximum response time or average delay). Althoughthese papers provide exact or numerical solutions tothe problem and give some important insights, theirassumptions often limit their applicability to real cases,as summarised in Table 1. Soman et al. (2004) assertthe insufficiency of mathematical approaches to solvethe decoupling point positioning problem, the mainreason being the difficulty in modelling the problemaccurately. For instance, many models in Table 1 donot consider set-up times or cost (Li 1992, Arreola-Risa and DeCroix 1998, Gupta and Benjafaar 2004,Kaminsky and Kaya 2008), or the possibility tostock intermediate items as an alternative to MTO orMTS for finished products (Li 1992, Arreola-Risaand DeCroix 1998, Rajagopalan 2002, Kaminsky andKaya 2008). Moreover, these models may: (i) rely ona complex procedure not easily understandable by asmall and medium sized enterprise (SME) supervisor(Gupta and Benjafaar 2004, Kaminsky and Kaya2008, Sun et al. 2008); (ii) require data difficultto collect or estimate (Gupta and Benjafaar 2004,Sun et al. 2008) or (iii) be based on simplisticassumptions about the number of products (Sunet al. 2008), the production system (Arreola-Risa andDeCroix 1998) or the demand process (Rajagopalan2002). Moreover, most of these models fail to providedetailed support for practical implementation: onlythe model by Rajagopalan (2002) is accompanied byan application to a real case.
Two papers in Table 1 propose qualitative frame-works, encompassing sets of decisions. Huiskonenet al. (2003) propose a four-step framework inwhich decoupling point determination is driven byproduct analysis (sales volumes and demand stability),complemented by customer analysis and grouping(on attractiveness and judgemental recommendationsabout the opportunity for maintaining, improvingor reducing the level of service). The outcome is aproduct/customer specific policy: therefore, the sameproduct might have different decoupling points for dif-ferent customers. Soman et al. (2004) indicate decouplingpoint setting as the first of three hierarchical decisionallevels, followed by production planning (due datepolicies for MTO products, lot sizes for MTS products,monthly production volumes) and scheduling and con-trol (daily/weekly production volumes and sequence).
560 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
Table
1.Summary
ofresearchondecouplingpointdecisionsin
hybridMTO–MTSsystem
s.
Paper
Approach
Main
contribution
Main
limitations
Analyticalmodels
Li1992
Analyticalmodel(foronefirm
andforn
firm
s)aim
edatprofitmaxim
isation.
New
sboy-likeform
ulasallow
theoptimalchoice
betweenMTO
andMTSoperationsandthe
optimalinventory
level
tobedetermined.
Single
product
system
.Set-upcostsnotconsidered.
Interm
ediate
productsnotconsidered.
Considerscost
structure
andspeedofdelivery
asadecisionaldriver.
Arreola-R
isaand
DeC
roix
(1998)
Analyticalmodel
aim
edatminim
ising
thetotalcost
(inventory
holding,
backordering).
Findsoptimalandexact
solutions.
Identifies
severaldecisionaldrivers(ratiobetween
unitholdingandbackorderingcost,capacity,
variabilityofprocessingtimes).
Single
stagemanufacturingsystem
.Set-upcostsnotconsidered.
Interm
ediate
productsnotconsidered.
Rajagopalan
(2002)
Analyticalmodel
minim
isingthe
inventory
holdingcost
ofMTSitem
s,subject
toaresponse
timeconstraint
forMTO
item
s.Case
studyofaplastic
productionfirm
with123products.
Findsheuristic
solutions.Proposesanem
pirical
application.
MTO
item
sperturb
MTSitem
production,increasing
congestion,leadtimeandinventory.In
thecase
analysed,verylowandveryhighdem
anditem
sare
MTO,theothersare
MTS.
Considersasdecisionaldrivers:dem
and,unit
processingtime,unitholdingcostandset-uptime.
Interm
ediate
productsnotconsidered.
Verylow
orsporadic
dem
andproducts
notconsidered.
Gupta
and
Benjafaar
(2004)
Analyticalmodel
minim
isinginventory
costsunder
amaxim
um
average
deliverydelayconstraint.
Solutionsare
given
throughnumerical
examples.Thecost
ofmodular
product
re-design(tohaveaunique
interm
ediate
forallitem
s)isalso
considered.
Analysestherelationamongcapacity
constraints,
decouplingthroughdelayed
differentiation
(interm
ediate
products),inventory
costandservice
level:in
general,thedesirabilityofdelayed
differentiationdependsontheamountofslack
capacity
available
(thelower,themore
aMTS
productionisfavoured).
Relatesdecouplingpointdecisionsto
product
design
decisions.
Themodel
hypotheses
makeitsuitable
forthe
PC
assem
bly
industry.
Decisionaldrivers:inventory
cost,backorder
cost,
product
re-designcost,timeperform
ance,
available
capacity
atdifferentproductionstages,
workload.
Set-upcostsnotconsidered.
Only
‘pure’MTSorMTO
situations
considered.
Themodel
usesaunitamortised
product/process
redesigncost
that
should
beestimatedforeach
possible
decouplingpoint.
Kaminskyand
Kaya(2008)
Analyticalmodel
minim
isinginventory
costs,leadtimeandtardinessin
the
whole
system
.Extensionto
atw
o-tieredsupply
chain.
Findsheuristic
solutions.Relatesthedecoupling
pointdecisionto
thejobsequencingrule
andthe
evaluationofwaitingtimes
asawayto
jointly
minim
isewaitingtimes
andinventory
holding
costs.Extendstheproblem
toasupply
chain
contextfortheevaluationofoverallcost
perform
ance.
Decisionaldrivers:timeparameters,cost
structure.
Set-uptimes
andcostsnotconsidered.
Interm
ediate
productsnotconsidered.
Verycomplexprocedure.
Complexdecisionalgorithm. (c
ontinued
)
Production Planning & Control 561
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
Table
1.Continued.
Paper
Approach
Main
contribution
Main
limitations
Sunet
al.(2008)
Analyticalmodel
minim
isingtheoverall
cost
(inventory,set-up,stock-outand
asset
specificity)in
asupply
network,
subject
tosatisfyingcustomer
delivery
time.
Applies
theproblem
toanassem
bly
rather
than
manufacturingcontext,within
asupply
network
wheremultiple
possible
decouplingpoints
exist.
Takes
into
considerationBOM
relationships
amongdifferentmaterials.Provides
numerical
solutions.Decisionaldrivers:componentand
product
cost
structure,dem
andvariance
and
customer
deliverytime.
Single
product
model.
Themodel
applicationrequires
the
estimationoftheasset
specificitycost
foreach
component.
Framew
orks
Huiskonen
etal.
(2003)
Framew
ork
inwhichproduct-based
inform
ationissupplementedby
customer-specificcharacteristics
when
decidingontheinventory
policy
fora
specific
product.
Case
studyofapackagingindustry
firm
with185productsand384customers.
Decouplingdecisionsare
relatedto
customer
aspects,
andare
madeforeach
product/customer
combinationonthebasisofquantitativeand
judgem
entalsteps.Decisionaldrivers:product-
basedinform
ation(dem
andvolumes
and
variation),customer
inform
ation(purchase
volumes,profitcontribution,growth
potential,
impact
ofservicelevel
oncustomer
volumes).
Nooptimisationapproach
todecision-
making.
Interm
ediate
productsnotconsidered.
Theproduct/customer
couplinganalysis
includes
judgem
ental(m
anual)steps:
itbecomes
verytimeconsumingfor
firm
swithawideproduct
rangeanda
largeportfolioofcustomers.
Somanet
al.
(2004)
Hierarchicalplanningframew
ork
developed
forthefoodprocessing
industry.
Points
outtheinsufficiency
ofmathem
atical
approaches
insolvingtheMTO–MTSdecision,
themain
reasonsbeingthedifficultyofmodelling
theproblem
accurately.Theconceptual
framew
ork
suggestedisanattem
ptto
structure
the
productionplanningdecisionsin
acombined
MTO–MTSproductionsituation.
Decisionaldriversconsidered:dem
andvolumes
and
variability,inventory
vs.set-upcosts,sequence-
dependency
ofset-upcosts,product
customisation
andperishability.
Interm
ediate
productsnotconsidered.
Theproposedframew
ork
provides
astructuredandhierarchicalsetof
decisionsto
bemade,
butonly
suggestionsonhow
differentdrivers
influence
thedecisionsare
provided.
562 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
Case
studies
VanDonk(2001)
Case
studyofafoodindustry
firm
buildinganew
facility,withmore
than200products.
Highlightsthecontrastingim
pact
ofdifferentmarket
andprocess
characteristics
onthedecoupling
pointin
thefoodindustry.Easily
understandable
bymanagers.
Decisionaldriversconsidered:(a)Product
and
market
characteristics:deliverytimeand
reliability,dem
andpredictabilityandspecificity;
(b)Process
characteristics:leadtime,coststructure
(includingset-up,control,stock
holdingand
obsolescence).
How
tosolvethetrade-offsisnot
explained.
Verylimited
quantitativeinsights
(even
inthecase
study).
Nooptimisationapproach
todecision-
making.
Theadoptedrulesfordecouplingpoint
decisionsare
case
specific.
Somanet
al.
(2007)
Case
studyapplyingtheframew
ork
of
Somanet
al.(2004)to
afoodindustry
firm
with230products.
Inorder
todefinethedecouplingpointforproducts,
adem
andvariabilityanalysis,andthen
aproduct-
process
analysisare
perform
ed.
Points
outtherelevance
ofqualitativeandshared
decision-m
aking:jointmeetingsofsalesand
productionpeopleshould
bearranged
toshowthe
resultofdetailed
analysisandto
discuss
the
implicationsofvariousscenarios.
Lim
ited
quantitativeinsights
onhow
toundertakethedecouplingpoint
positioning:analyticaldecisionaids
are
provided
only
forscheduling
problems.
Kerkkanen
(2007)
Case
studyofasm
allMTO
steelmill
planningto
movetowardsahybrid
MTO–MTSsystem
.
When
alargeamountofinform
ationthatisneeded
formakingthedecisionisin
anintangible
and
qualitativeform
,organisationalfactors
will
dominate.
Thereisaneedto
combinequalitativeand
quantitativeinform
ationandto
increase
communicationsurroundingcriticalissues.
Thedecisionprocess
illustratedis
specific
totheproblem
andcase
companyanalysed.
Thefocusofthepaper
isonfinding
factors
affectingcommonality
among
thepotentialinterm
ediate
item
s(i.e.
grades
ofsteel,billetdim
ensions,
dem
and),rather
thanonanalysingthe
MTO–MTSissueatastrategic
level.
Production Planning & Control 563
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
Their conceptual framework can be used by managersas a starting point for designing or redesigning theplanning and scheduling hierarchy structure for aparticular situation. The framework, especially devel-oped for the food processing industry, is appliedto a case study company, as described in Somanet al. (2007).
The referred frameworks have advantages overanalytical models, as they are more easily understoodby managers and easier to implement in real cases.On the other hand, they do not lead to optimalsolutions for the cost-service trade-off, and theyprovide suggestions or guidelines rather than detailedprocedures or quantitative decision aids. Finally, casestudies such as those by Kerkkanen (2007) and VanDonk (2001) are rich in interesting insights arisingfrom the company analysed, but the decision problemis solved through very specific rules or procedures,which are difficult to generalise beyond the specificcontext considered.
The papers in Table 1 help to identify a list offactors that can be considered as decisional driverswhen setting the decoupling point. One class of driverscorresponds to the product cost structure, encom-passing inventory holding, backordering, set-up, con-trol and obsolescence costs, and several ratios amongthem. Products with high inventory and/or obsoles-cence costs are more suitable for MTO production,while MTS production best suits products affectedby relevant backordering, set-up and control costs.A second category concerns time parameters, includinglead time, delivery time, processing times and theirvariability and/or predictability: long and unpredict-able times require companies to buffer demand withstock (i.e. MTS production) while MTO productioncan suit products with short and more reliable times.A third category of drivers concerns demand in itsabsolute volume, variability over time, specificity andpredictability: the more one product is affected bya high, regular, general and predictable demand,the more it makes sense to fulfil it from stock, andvice versa. A further relevant aspect in some sectors isproduct ‘perishability’, as Soman et al. (2004) reportfor food processing.
Most of the aforementioned papers acknowledgethat these characteristics are related with the decou-pling point positioning. Yet some papers underlineother aspects as well. For instance, both Arreola-Risa and DeCroix (1998), and Gupta and Benjafaar(2004) refer to the relationship between availableshop floor capacity and demand and its effect on thepositioning of the decoupling point. Furthermore,Soman et al. (2004) consider the sequence-dependency
of set-ups as a relevant factor. It is suggested that
when a company is short of capacity, it should
resort to large lots (and therefore MTS rather than
MTO production) in order to free capacity by
reducing machine time lost to changeovers; however,
when set-up times are sequence-dependent, a set-up
time reduction could also be achieved by means of a
sensible dispatching policy, as noted by Perona and
Portioli (1993).Even the product structure and bill of materials
(BOM) could influence whether an item is most
suitable to be produced to order or to stock, as noted
by Sun et al. (2008). In assembly systems, common
components or subgroups might be kept in stock
and assembled to order, and much the same might
happen in manufacturing when a semi-finished item
can evolve into several different finished products.
Finally, Huiskonen et al. (2003) underline that spe-
cific customer requirements could dictate that the
same product is produced differently for different
customers.Overall, the above discussion shows that decou-
pling point decisions have not been adequately
addressed in the available literature, and points to
limitations in theoretical approaches, as well as to the
scarcity of practical applications. On the one hand,
analytical approaches use rational and quantitative
models, but do not seem suited for being easily
implemented in practice, and actually have not been
implemented (with one exception). On the other hand,
case studies show that, in reality, managers tend to
base their decision-making mostly on qualitative and
intangible aspects. This is the theory–practice gap first
noted by Van Donk (2001), for which several reasons
exist. As pointed out by Kerkkanen (2007), inventory/
production decisions have cross-functional effects
on functions that have different levels of knowledge
of the problems treated. Moreover, SMEs might
find it difficult to deploy the (financial and human)
resources needed to implement complex decision
models. Finally, rational and formal models do not
take into account the effects of most contextual
factors that accompany decision-making in real-world
settings (Cohen 1993). Therefore, experienced decision-
makers operating in their ‘natural’ environment are
‘more concerned about sizing up the situation and
refreshing their awareness through feedback, rather
than developing multiple options to compare one another’
(Zsambok 1997). Consistent with this, they tend to
base their decisions on experience and common
sense. All these observations call for approaches
that are simple and illustrative, without giving up the
564 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
ability to support the decision-making process in arational way.
3. Research methodology
To bridge the theory–practice gap, the research couldeither start with a theoretical model and refine it tobetter suit practice or vice versa. The first, ‘technologypush’ approach, might be the best choice when theorybuilding is the main objective, as argued by Stevensonand Silva (2008). In contrast, the second ‘marketdriven’ approach seems better suited to pursue thespecific objectives of this study. This article, in fact,marks the first step in a wider research program whichaims to design a theoretically sound planning frame-work through a refining process that starts withpractice, by adding to and/or improving rational andquantitative decision-making procedures. Consistently,our article does not propose any new planning policy:rather it drafts some new ideas on how establishedpolicies should be selected, combined and assignedto various products in order to fit their most relevantand sometimes conflicting requirements. This articleonly considers inventory planning issues: productionand capacity, as well as purchasing, are left forfuture development. The new framework proposed inSection 5 has been put into practice in the companydescribed in the next section.
Although our new framework does not pursue anyexplicit objective function, its general objective can bepractically defined as the search for inventory policiesthat are able to reduce the total inventory and set-upcost, without deteriorating (or better, improvingat the same time) the customer order response timeand the timeliness performance. Thus, in order tomeasure the potential of our framework, cost, time andtimeliness performances achieved under the old deci-sional system had to be compared with those gatheredusing the new decisional framework. Despite theavailability of results achieved by using the newplanning framework in practice, we considered thatthe comparison of these data with the old approachwould not provide a clear indication of the newframework’s potential, since the new data stem notonly from new methods but also from new anddifferent demand, and are also affected by the newinformation system and a new production manager.As a consequence we performed a static simulationanalysis based on a spreadsheet, so that the old andnew approaches could be compared under the sameconditions. Historical data on customer and produc-tion orders for the past 3 years were used as the
starting point. We divided this time frame into twointervals: an ex-ante period, consisting of the formertwo years, which was used for the item segmentation,planning policy setting and parameters calculation,following the framework described in Section 5, and anex-post period, corresponding to the last year, used toperform the static simulation of the new decisionalenvironment (NEW situation) and the comparisonwith the historical performances (OLD situation)achieved in the same year.
The direct output of the ex-post period simulationwas the entire set of production orders issued foreach product. Thus, the performances that could becomputed and used as key indicators in the compar-ison of the two planning environments are: theaverage lot size; the average stock; the average delayin customer order fulfilment; the set-up costs incurredin the planning horizon; and the holding costs inthe planning horizon. The performances achievedby the new planning environment were assessed byfirstly simulating them for a sample of representativeproducts, and then comparing them with the corre-sponding performance actually achieved by the sameproducts in the same ex-post year, but with the olddecisional context. Starting from the comparativeperformances achieved by the sampled items, weextrapolated the whole production mix behaviour byextending the average advantage achieved by thesampled products to all of the products, weightedbased on the demand volume of each product.Although this might not be the most scientificallyrigorous method to do this comparison, we foundit to be the most practical given the vast amountof products that should be evaluated (Section 4).Moreover, several cross-checks confirmed the overallsoundness of projected results.
4. The case study company
The company considered in this study manufacturessteel wires. It is a small–medium sized enterprise,employing more than 100 people and with annualrevenues of approximately E35 million. The productrange consists of more than 3000 finished products,differing by: chemical composition, diameter andtolerance, mechanical characteristics, surface finishingand packaging. The overall production volume is inexcess of 13,500 tons of finished wire per year. Thenumber of active customers is more than 300, 13% ofwhom account for 80% of the overall turnover.Demand is irregularly distributed: while a smallrange of standard products have a rather high and
Production Planning & Control 565
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
predictable demand, many other products experience
an unpredictable and even ‘lumpy’ consumption.The manufacturing process consists of the
following phases, illustrated in Figure 1:
(a) An initial surface treatment prepares the
material for the following drawing phase;(b) The wire drawing is performed until the target
diameter is reached;(c) An annealing treatment and, if needed;(d) other surface treatments, allow the required
strength and tolerance to be obtained;(e) The final activities are quality control and
packing.
If a very high diameter reduction is needed, the sequence
(surface treatment! drawing) can be repeated up to
three times until the required diameter is obtained.Cycle routings may vary significantly, as a function
of: the diameter reduction ratio, the type of material,
and the required mechanical characteristics. Machines
incur set-up activities that may last from 20min to
several hours according to the product sequence.
Given the extremely wide product range supplied,
the ‘default’ demand fulfilment policy adopted was
MTO. By exception, some particularly high and regular
demand products were produced for stock, with a target
stock level issued per week for each item and their
availability updated and faxed to key customers.No formalised or quantitative criteria were
adopted to support decision-making: for instance,
items considered appropriate for MTS production
were identified based on common sense and experi-
ence, rather than by numerically examining demand
patterns. Likewise, lot sizes were set by considering
the space available in the warehouse, as well as the
raw material coil dimension; the classical trade-off
between inventory and set-up costs was not considered.
Moreover, MTS orders were issued to keep machines
running rather than to replenish low inventories.No specific, formalised tool for inventory planningwas established within the company’s informationsystem. As a result, delivery times were long andunpredictable (ranging from 5 to 80 days), with a pooron-time delivery record. Many orders were deliveredearly and many others delivered late; the averagelateness was almost 6 days.
5. The new framework
5.1. Overview
Given the specific objectives pursued in this article, ouraim is not to suggest new inventory policies but ratherto devise a new framework to select and assign existingpolicies to the different products of the company toimprove efficiency (inventory and set-up costs) andeffectiveness (delivery performances). The new frame-work proposed in this article consists of four maindecisional steps:
(i) Products are segmented into homogeneousgroups with regard to demand volume andspecificity to customers (Section 5.2).
(ii) Next, an appropriate decoupling point isdetermined for each group on the basis ofdemand and customer analysis, as proposed inHuiskonen et al. (2003) and Soman et al. (2007),by considering the option of delayed differen-tiation as in Gupta and Benjafaar (2004).
(iii) Then, a specific replenishment policy is assignedto each product group on the basis of specificcustomer requirements (Section 5.4).
(iv) And finally, the parameters (lot sizes andtarget stock availability) needed to apply eachreplenishment policy to each considered prod-uct are computed by considering set-up andinventory holding costs and time parameters(Section 5.5).
Figure 1. Scheme of the production process in the case study company.
566 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
Table 2 gathers the link between each frameworkstep and the existing literature.
Some of the decisional drivers identified in theliterature review described in Section 2 were notconsidered. For example, given that the companymanufactures steel wiring, perishability was not arelevant issue. Also, we did not have to consider therelationship between demand and available shop floorcapacity, since the company did not experience capac-ity constraints. However, capacity should be consid-ered in later versions of our decisional framework inorder to make it more general. Likewise, although set-up times for drawing benches are (moderately) affectedby the production sequence, this was not considered.In fact, given the medium term horizon consideredby our framework, we decided for the sake of sim-plicity to base computations on an average, sequence-independent set-up time.
5.2. Step 1: segmentation into homogenousproduct groups
This first decisional step identifies groups ofhomogenous products with regard to their sugg-ested decoupling point and inventory planning policy.It is performed through Pareto multi-criteria analysis(Flores and Whybark 1986), combining physicalvolumes with the number of customer orders(according to the two digit ABC analysis proposedby Krupp (1994)). The cross-analysis structure isshown in Figure 2.
The analysis generates five groups. Group 1consists of finished products with a very high andregular demand, resulting from the combination ofmany small orders issued by a large number ofdifferent customers. Group 2 encompasses finishedproducts with an almost equally high demand, but onlya few large orders: they are special products custom
made in large quantities for specific large scale
customers. Finished products with a medium level of
demand and number of orders are classified as Group
3. The BOM for the remaining (and very low demand)
finished items is analysed in order to identify common
intermediates. Intermediates can be examined using
the same bi-dimensional Pareto analysis as for fin-
ished products (production volumes and orders are
considered instead of sales and customer orders).
The common intermediates with a large and regular
demand, are assigned to Group 4, while the remaining
finished products (with no common intermediates) are
placed in Group 5.The segmentation process is based on demand
history, while a new product would fall under one of
the following three scenarios: (a) it is phased-in as a
variant of a pre-existing product that is phased-out
accordingly: in this case the new product is inserted
into the same group as the phased-out old product;
(b) it is introduced to fulfil a custom request of a
large scale customer with a shared order portfolio,
covering several months or even years: in this case
it fits the requirements of Group 2; (c) otherwise, the
Table 2. Link between the steps of our framework and existing literature.
Framework step Related literature Considered aspect
Productsegmentation
Flores and Whybark (1986)Krupp (1994)Huiskonen et al. (2005)
Multi-criteria Pareto analysis
Decoupling pointdetermination
Huiskonen et al. (2003)Soman et al. (2007)
Decoupling point set on the basis of demandand customer analysis
Gupta and Benjafaar (2004) Delayed differentiation option considered(e.g. MTS intermediate items)
Inventory policyassignment
Silver et al. (1998) Criteria for policy assignment
Parameter setting Silver et al. (1998) Basic criteria for parameters settingsSilver and Robb (2008) Advanced criteria for parameters settings
Figure 2. Pareto cross matrix for finished products andintermediate items.
Production Planning & Control 567
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
new product is firstly inserted into Group 5. Sincethe whole procedure is periodically reiterated in orderto refresh the various decisions on the ground ofnew data, the assignment of new products can bere-assessed as soon as sufficient historical demand datais available.
5.3. Step 2: decoupling point determination
Three possible decoupling points are considered in ourframework (as exemplified in Figure 1): raw materials,intermediates and finished products in the warehouse.In accordance with Silver et al. (1998), products withingroups 1 and 2 are MTS. Therefore, their decouplingpoint was positioned at the finished product level.Products belonging to Group 4 are decoupled at theintermediate level. Many products in the group sharethe same intermediate: as a consequence, althougheach finished item has low and unpredictable demand,the corresponding intermediates share a rather highand regular production volume. This way of fulfillingcustomer demand was named make-to-stock-finish-to-order (MTSFTO): for further details on the finish-to-order policy, see Dobson and Stavrulaki (2003).Finally, the decoupling point for items belonging togroups 3 and 5 is positioned at the raw material level(i.e. MTO), given their unpredictable and irregulardemand. This is logical because each raw materialcan be used to manufacture on average around 50–60finished products, so all raw materials tend to have apredictable and regular demand pattern.
5.4. Step 3: replenishment policy assignment
Table 3 illustrates the replenishment policies assignedto each group, combined with the respective decou-pling point.
An MTS policy was considered appropriate forproducts in group 1, characterised by regular andrepetitive demand consisting of many small ordersissued by several independent customers. The replen-ishment policy proposed is a variation of the (R, s, S)that Scarf (1960) proved, under quite general assump-tions, to minimise the total review, replenishment,carrying and shortage costs. More precisely, thereplenishment policy selected for items within Group1 is a fixed reorder time model (R, S) (Silver et al.1998). Replenishment orders are released at a fixedtime interval (R), with a quantity that allows the totalavailable stock to be restored to the order-up-to level(S). In accordance with Silver et al. (1998), the safetystock was set so as to guarantee a given fill rate and tocover demand variability during the replenishment leadtime. The simple periodic stock review required inorder to implement this inventory policy was sup-ported with a semi-automated spreadsheet procedure.A more detailed discussion on the (R, S) periodicreview inventory system and its optimality conditions,together with some managerial interpretations, can befound in Silver and Robb (2008).
Products in Group 2 are characterised by a highand regular demand and are purchased by only onecustomer: hence, customer–supplier collaboration iskey for inventory planning. Customers should providea periodic consumption plan, with a single item detailover a horizon of some periods, with different levelsof flexibility on the quantities indicated. A rollingcollaborative requirement planning policy has tobe activated: customer demand plans allow the netrequirements of finished products to be computed,and production and raw material replenishment ordersto be issued.
Products within Group 3 have medium sales withsome regularity. Demand is not high enough to justifya pure MTS policy, while the existence of severalcustomers and orders for these items suggested thatreplenishment order quantities could be planned muchlarger than the average customer order, to save onchangeover time and cost. Therefore, production ofthese items is triggered by a customer order, but theminimum production quantity has been set to theeconomic order quantity (EOQ; see Harris (1913)). Asa consequence, each time a replenishment order isissued, part of it serves the related customer orderwhile the remainder is stored and used to meet futurecustomer orders. Group 4 encompasses semi-finisheditems with high and regular volumes, used for a largenumber of finished products and managed using amake-to-order policy. Consistently, their decouplingpoint is set at the intermediate item level. Thereplenishment policy chosen for intermediate items
Table 3. Inventory policies combined with decoupling pointpositioning.
Decoupling point positioning
Replenishmentpolicy
Rawmaterials Intermediate Finished
L 4 L Group 5 Group 4(intermediateto finished)
–
EOQ Group 3 –Monthlyrolling
Group 2
R, S Group 4 (rawto intermediate)
Group 1
568 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
within this group is similar to that chosen for finisheditems within Group 1, since the demand characteristicsof these two groups match. Moreover, the correspond-ing finished items are replenished in a lot-for-lot (L4L)fashion, given the absence of repetitiveness in theirdemand patterns. Finally, products in Group 5 arealso replenished by means of a L4L policy becauseof their very low and sporadic demand. Therefore,these products can be produced only on the basis ofcustomer orders, and production matches the customerorder quantity.
5.5. Step 4: parameter setting
The setting, monitoring and periodic updating ofparameters can be supported by a simple informationtechnology tool, which gathers data from a centraldatabase, loads it into a preconfigured spreadsheet andenables the parameters for each product in each groupto be updated. Given the inventory policies describedabove, and according to Silver et al. (1998), thefollowing parameters have to be set:
(i) The EOQ, used both directly in Group 3 toset the replenishment quantity and indirectlyin groups 1 and 4 to set the review period (r).The EOQ is calculated according to theHarris–Wilson EOQ formulation (Harris 1913, Wilson1934) that requires three main data, as follows:
. Expected demand, which is extrapolated onthe grounds of historical demand.
. Set-up costs, that consider both physicaland opportunity components. The oppor-tunity cost connected with the capacity lossdetermined by each set-up is computedas the contribution margin. The physicalset-up cost is, instead, the sum of all directcosts connected with changeover activities,such as specific materials.
. Holding costs, computed as the sum ofa financial and a physical component.The financial part depends on the need toanticipate the working capital connected toraw material, WIP and finished productsstill on hand. It is evaluated by takinginto account the marginal interest rate onfinancial debt. The physical storage costis evaluated on the basis of a third-partywarehousing facility pricelist.
(ii) The review period, R, used to schedule stockreviews and the issuing of replenishmentorders in groups 1 and 4 was set equal to thenumber of weeks that best approximates
the EOQ model under average demandconditions.
(iii) The order-up-to level, S, used to determine thereordering quantity in Groups 1 and 4, was setto the sum of the average demand during thereview period and production lead time plussafety stock (computed according to the vari-ability of the demand during the lead time).
6. Empirical application
6.1. Projected results
The framework illustrated in Section 5 was appliedto the case company described in Section 4. The resultsof product segmentation and the related inventorypolicies assigned are reported in Figure 3. The smallnumber of rather repetitive products clustered inGroups 1, 2 and 3 produce almost 70% of thevolume and sales despite accounting for less than15% of products. They define a ‘standard and repet-itive factory’ within the overall factory in contrast tothe ‘custom and lumpy factory’ encompassing the pureMTO products in Groups 4 and 5. Products belongingto the ‘standard and repetitive factory’ should bemanaged with a focus on efficiency, while effectiveness(achieved with a short delivery time) should be consid-ered as given. By contrast, custom products should bemanaged by focusing on effectiveness (especially designquality, conformity and delivery time).
In order to evaluate the benefits that can beachieved with the framework application by the casecompany, we performed a simulation of the newinventory policies under the ex-post period of oneyear, as described in Section 3. In the following,we report the results of the comparison between theold and new decisional context for three products,one from each of Groups 1, 2 and 3, chosen for being
Figure 3. Product segmentation results.
Production Planning & Control 569
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
particularly representative of their groups. It wasnot possible to evaluate Group 4 because no histori-cal data was held in the old information systemabout intermediate items, while the inventory policyfor Group 5 did not change substantially and soa comparison was not necessary. Table 4 presentsthe exogenous and endogenous parameters for theselected products while the comparison is reportedin Table 5.
Table 5 reveals how the new planning frameworkdramatically reduced the number of production orders,yielding an increase in average stock and correspond-ing holding costs. However, this increase is more thanoffset by the reduction in set-up costs, meaning totalcosts are also reduced.
Note that holding costs in the NEW scenarioare computed considering a third-party warehousingservice, so as to take into account the overall inventorycost regardless of the amount of additional spacerequired. This explains the difference between theaverage increase for the quantity in stock (þ40%) andthe increase in holding costs (þ200%). For timeliness,items in groups 1 and 2 (accounting for 63% of theoverall production volume) are supplied from stockunder the NEW scenario, so their delivery time anddelay are equal to zero. Moreover, items in Group 3
can be supplied with a much reduced delay in theNEW scenario, given the lot sizing policy adopted. Thesample item chosen in this group had 16 customerorders in the considered ex-post year: in the OLDscenario as many as 11 production orders were issued,meaning that only 4 out of 16 customer orders(25% of the total amount) were fulfilled from stock.In the NEW scenario, only 4 production orders wereissued to cope with the same number of customerorders: therefore the percentage of customer ordersfulfilled from stock has increased to 75%, whichexplains the average delay reduction.
Results aggregated and projected (as explained inSection 3) are reported in Figure 4, where a decrease inthe total cost is achieved due to the balancing effect ofset-up and holding costs that leads to an overall savingin excess of 36%. According to historical data, themain planning policy used to be an almost pure MTO,which is confirmed by a very low level of holding costs,and by set-up costs which account for more than92% of the total costs within the old decisionalenvironment. However, this almost pure MTO policyis significantly outperformed by the mixed policiesconsidered by the new decisional context, because ofthe better balance achieved between inventory holdingand set-up costs.
Table 5. OLD vs NEW performance comparison for the three selected items.
Group 1 Group 2 Group 3
Parameter OLD NEW D% OLD NEW D% OLD NEW D%
Number of production orders 22 5 �77 21 11 �48 11 4 �64Average production order [ton] 1.9 11.8 þ520 4.6 9 þ95 3.1 8.4 þ171Average stock [ton] 5 7 þ40 1.63 1.27 �22 0.8 2.8 þ250Average delay [days] �0.9 (max¼ 49!) 0 �100 3.1 0 �100 4.2 2.3 �45Set-up costs [E/year] 4044 920 �85 928 486 �48 2021 735 �64Holding costs [E/year] 486 1456 þ200 162 126 �22 144 504 +250Total cost 4530E 2376E �48 1090E 612E �44 2165E 1239E �43
Table 4. Product parameters.
Representative items from each group
Group 1 Group 2 Group 3
Demand 63 [tons/year] 100 [tons/year] 34.6 [tons/year]133 [customer orders] 23 [customer orders] 16 [orders]59 customers 1 customer 4 customers
Planning policy Review period¼ 9 [week] N/A EOQ¼ 8.4 [tons]Reorder level¼ 14.6 [tons]Safety Stock¼ 3.4 [tons]
570 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
Note that the results for the NEW decisionalenvironment for products within Groups 4 and 5 wereassumed to be unchanged from the OLD environ-ment. While this is reasonable for products in Group5, it does not account for the reduction in set-up andinventory costs gained by shifting the decouplingpoint for Group 4 products from the raw materialto the intermediate product stage. Moreover, stockholding costs within the NEW decisional scenariowere computed by considering the prices of anexternal warehousing facility, given the sharp increasein inventory suggested by the new framework: never-theless it was later discovered that enough space wasavailable to cope with the increase in inventory (seeSection 6.4. for more details). Thus, results presentedin Figure 4 are likely to underestimate the overall costreduction obtained.
6.2. Implementation issues
Once the benefits achievable thanks to the newdecisional environment were estimated through simu-lation, the company decided to implement the newdecisional framework in practice. The new planningenvironment has been transferred to the companymanagers and supervisors (especially in sales andproduction). The transfer process encompassed thefollowing steps, which took approximately 6 monthsto complete:
(i) First, the various decisional drivers identifiedin the literature were introduced and dis-cussed, in order to find the best fit with theenvironment.
(ii) Then, the segmentation methods proposedwere explained to managers and supervisorsand an appropriate means of implementingthem was explored.
(iii) Next, the choice of decoupling points andinventory policies was discussed in depth. Forproducts in Group 2 a monthly plan with afour month horizon was arranged, with dif-ferent levels of flexibility allocated to differentmonths: quantities forecast for month 1 wereconsidered as fixed orders; quantities forecastfor month 2 were allowed a review margin of�10%; and quantities indicated for months 3and 4 were considered as only indicative andopen to change by the customer. Moreover,the MTSFTO policy selected for productsand intermediates within Group 4 requiredBOM to be updated. This was a long anddelicate task.
(iv) Then, we developed a set of spreadsheetapplications, in order to download customerorders and technical data from the centraldatabase and to make all the computationsneeded in order to segment products anddetermine their inventory policy parameters.
(v) And finally, these applications were trans-ferred to managers. At the end of the trans-fer process, the staff involved could use thespreadsheet-based applications and all appre-ciated their ease of use and transparency.
As for the results, it is difficult to compare themto any of the performances illustrated in Section 6.1because they were not only due to the new planningdecisional framework but also to the different decis-ional style of a new production manager, the differ-ent informational and decisional support of the newinformation system, and (above all) a new set of inputdata (for example, a different production mix withdifferent customer orders). Nevertheless, in the yearfollowing our project, the company’s sales soared by20%, due to the combined effect of an increase in theprice of finished products (due in turn to an increasein the price of steel, and a shift in the production mixfrom carbon to stainless steel wire) and an increasein overall output. An increase in output was facilitatedby more capacity at the drawing benches as a resultof fewer changeovers, made possible by larger lot sizescombined with the switch from pure MTO to someforms of MTS.
Meanwhile, the quantity of finished productsin stock more than doubled because of the largerproportion of products and volumes that were pro-duced to stock. The availability of some free spaceallowed the additional inventory to be stocked withinthe plant. Moreover, the warehousing of finishedproducts was further simplified by the adoption ofthe new decisional framework due to the sharp
Figure 4. Cost comparison generalisation.
Production Planning & Control 571
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
reduction in the number of handling operationsconnected with larger production lots. However, thiscost reduction has not been quantitatively evaluatedand considered in Figure 4 because both personnel andequipment remained within the company and were notoutsourced.
7. Discussion and conclusion
In Section 5 we presented a new, easy-to-use andsufficiently straightforward decisional framework toplan inventory in a generic multi-stage manufacturingcompany. We described a field implementation as ameans of developing a theoretically sound, yet practi-cally usable, method. The test, performed at a small steelwire producer, supplied evidence that the new frame-work can provide a significant performance improve-ment when compared with a less formalised planningapproach. An improvement in the trade-off betweeninventory and set-up costs was achieved, improvingon-time delivery performance at the same time.
The reasons why this positive result was achievedare threefold. Firstly, this method addresses a hierar-chical gap in the operations decisional frameworkof the targeted company. Before the implementationof the new method, decisions tended to be taken eitherat the long-term (budget), or short-term (scheduling)level, while the mid-term (planning) level was almostdisregarded: this led to a poor alignment betweenlong and short term plans and resulted in an ‘all isurgent’ syndrome. Secondly, the framework identifiedtwo different ‘virtual factories’ co-existing within thesame manufacturing system. Products with a high andsteady demand are few but represent a large percentageof volume and sales. By offering these to customersfrom stock (or by producing them in larger lots),not only has the company achieved an unprecedentedlevel of efficiency, but also has the opportunity tofulfil demand that previously was lost to competitors.Improving efficiency also freed up capacity that couldbe used in a reactive manner when producing otherproducts to order or when attempting to accommodatelarger production volumes. Thirdly, planning decisionsare now based on logical and quantitative reasoning,supported by formalised methods and tools, and aimedat improving rational and quantitative objectives.The lack of formalised and straightforward planningactivities, together with the tendency to use the sameplanning policies for all items and products simplybecause they are produced in the same way, isconsidered commonplace in many manufacturingSMEs. Hence, the proposed method could also beuseful in other similar companies.
It should be noted that the projected results showa severe increase in holding costs (þ200%); this isconsidered appropriate for the particular setting ofthe case study selected but the proposed frame-work does not necessarily lead to an inventoryincrease or decrease. Rather, it attempts to improveoverall performance by better balancing inventoryand changeover costs. In line with this remark,Cannon (2008) underlines that inventory reductionis not always a solid indicator of overall improvedperformance. It is also important to note that set-upcost savings are connected with the sharp reduction inchangeover operations obtained by setting largerproduction lots. As explained in Section 5.5, part ofthe set-up costs was computed as the opportunitycost connected with the loss of production capac-ity consumed by changeover operations. Therefore,a part of set-up savings achieved refers to poten-tial savings in lost capacity (a detailed discussion on thiseffect can be found in Spence and Porteus (1987)).
This article adds to the available literature bycontributing to bridging the theory–practice gap,proposing a rational and quantitative inventory plan-ning approach which retains its usability in practicalenvironments. We chose to ground our study in prac-tice and proceed towards theory, instead of doing thereverse. Thus, the new decisional framework proposedwas developed using standard, well-established andeasy-to-use methods, requiring only simple and readilyavailable data and being supported by standard tools.Nonetheless, the proposed approach incorporatesseveral contributions from the theoretical literaturediscussed in Section 2. In particular, our frameworkadopts decisional drivers proposed in previous works,as highlighted in Section 5.1: the demand and customeranalysis, the delayed differentiation option, and thecost structure. However, in Section 2 we pointed outthat the analytical models in the literature have limita-tions that prevent real-life application. Comparing ournew approach with the six analytical models analysedin Section 2, we believe that our framework overcomesmost of their limitations, in that:
(i) it handles, without major problems, a realenvironment with fivemajor production stages,three different decoupling points and morethan 3000 final products. On the contrary, thereviewed models are single-product (Li 1992,Sun et al. 2008), single-stage (Arreola-Risaand Decroix 1998) and/or do not considerintermediate products (Rajagopalan 2002,Kaminsky and Kaya 2008);
(ii) it considers set-up costs as a relevant decisio-nal driver, accommodating decision support
572 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
for products with high and stable demandalongside products with sporadic and irregulardemand. Set-up costs are neglected by four outof six models analysed, and low demand itemsare not considered in some other models(Rajagopalan 2002);
(iii) it accommodates the choice of a differentdecoupling point, replenishment model andconnected parameters for each finishedproduct, intermediate item or raw material.Models in the literature allowing for a decou-pling point at an intermediate level consideronly pure MTS or pure MTO inventorypolicies (Gupta and Benjafaar 2004), or over-simplify the product range to a single-productcase (Sun et al. 2008);
(iv) finally, and perhaps most importantly, it doesnot require a PhD laureate for it to be imple-mented and run. Most analytical models in theliterature are arguably too complex for an SMEproductionmanager, or require data difficult tocollect or estimate, as highlighted in Section 2.
Although our new framework, differently from theanalytical models in the literature, does not optimiseany explicit objective function, we find that it supportsa large amount of decision-making with quantitativeand rational methods as compared to the otherframeworks and case studies reported in Section 2.In fact, the cost structure as well as time performances(timeliness of deliveries) are quantitatively assessed as away to define the thresholds for product segmentation(and therefore the decoupling point and inventorypolicy decisions) through a trial-and-error approach,in order to find a ‘good compromise’ between costsand service level. Moreover, since its quantitativemethods can be supported by automated and easy-to-use tools, the application of our framework mightimply less judgemental and time-consuming steps thansuggested, for instance in the case studies presentedby Van Donk (2001), Huiskonen et al. (2003) orKerkkanen (2007). In our case application, mostjudgmental analysis on the product segmentationresults was devoted to Groups 1 and 2, whichaccounted for a small fraction of the whole productmix. Furthermore, our approach suggests a rathergeneral framework that could be applied to severalindustrial contexts, while part of the analysed literaturecalls for an application to a very specific context, suchas the food industry (Van Donk 2001, Soman et al.2004, 2007) or a steel mill (Kerkkanen 2007).
The proposed approach has some limitations thatfuture research should address. Firstly, our decisionalframework does not consider such relevant planning
activities as capacity and purchasing: these were notcritical tasks within the considered case study butshould be added before our method can be consideredsufficiently general, as suggested by Arreola-Risa andDeCroix (1998) and Gupta and Benjafaar (2004).Secondly, despite the fact that the decisional frame-work is generally more rational and quantitative thanthat previously used in the considered company, wewould like to increase its quantitative and rationaldecision-making support. For instance, it would beinteresting to formulate an explicit optimal objectivefunction for costs or profit; an equally interesting newfeature would be to set thresholds for the segmentationmatrices depicted in Figure 2 in a way that optimisesthe objective function – this would be an improvementon the current trial-and-error approach. Thirdly, manymore real-life cases should be tested before any suchmethod can be declared generally applicable. Andfinally, although this research program aims to developpractically usable methods, quantitative comparisonsshould be made with scientific methods found in theliterature before we can be satisfied with the resultsachieved.
Acknowledgements
The authors wish to thank the (anonymous) referees for theirdemanding comments that indicated how to obtain, in ourview, a sharp increase in the quality of this article.
Notes on contributors
Marco Perona is full professor ofIndustrial Logistics at the Universityof Brescia (Italy). His research activityis devoted to operations management,supply chain management, IT man-agement and service management.In these fields he has published todate more than 40 papers in highlyregarded international journals. He is
Scientific Director of the Research Centre on ServiceManagement at the University of Brescia. He leads a smallteam of researchers and consultants with a portfolio ofprojects within leading international companies such asNestle, Procter & Gamble, Kone, Mediamarket and Ferrari.
Nicola Saccani (Department ofMechanical & Industrial Engineering,University of Brescia, [email protected]) holds a PhD inOperations Management, and isAssistant Professor at the Universityof Brescia (Italy). His research andpublications concern service opera-tions and management, inventory
Production Planning & Control 573
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
management, and buyer–supplier relationships. He isa member of the Research Centre on ServiceManagement at the University of Brescia, and of the ASAPService Management Forum initiative (http://www.progettoasap.org/).
Simone Zanoni ([email protected])is currently Assistant Professor at theMechanical & Industrial EngineeringDepartment at the University ofBrescia (Italy). He graduated (withdistinction) in 2001 in MechanicalEngineering and took his PhD atthe University of Brescia (Italy) in2005. His main research interests are:
Layout Design, Inventory Management and Closed-loopSupply Chains.
References
Arreola-Risa, A. and Decroix, G.A., 1998. Make-to-orderversus make-to-stock in a production-inventory system
with general production times. IIE Transactions, 30,
705–713.Cannon, A.R., 2008. Inventory improvement and financial
performance. International Journal of Production
Economics, 115, 581–593.Cohen, M.S., 1993. The naturalistic basis of decision biases.
In: G. Klein, J. Orasanu, R. Calderwood and C. Zsambok,eds. Decision-making in action: models and methods.
Norwood, NJ: Ablex Publishing, 51–99.Dobson, G. and Stavrulaki, E., 2003. Capacitated, finish-
to-order production planning with customer ordering day
assignments. IIE Transactions, 35 (5), 445–455.Flores, B.E. and Whybark, D.C., 1986. Multiple criteria
ABC analysis. International Journal of Operations &Production Management, 6, 38–46.
Federgruen, A. and Katalan, Z., 1995. Make-to-stock ormake-to-order: that is the question; novel answers to
an ancient debate. Working paper, Graduate School of
Business, Columbia University, New York.Federgruen, A. and Katalan, Z., 1999. The impact of adding
a make-to-order item to a make-to-stock productionsystem. Management Science, 45 (7), 980–994.
Gupta, D. and Benjaafar, S., 2004. Make-to-order, make-to-stock, or delay product differentiation? A common
framework for modeling and analysis. IIE Transactions,
36, 529–546.Harris, F.W., 1913. How many parts to make at once?
Factory: The Magazine of Management, 10, 135–136 ,
(Republished in Operations Research, 38 (6), 947–950,1990).
Hoekstra, S. and Romme, J., eds., 1992. Integral logisticstructures: developing customer-oriented goods flow.
London: McGraw-Hill.Huiskonen, J., Niemi, P., and Pirttila, T., 2003. An
approach to link customer characteristics to inventory
decision-making. International Journal of Production
Economics, 81–82, 255–264.Huiskonen, J., Niemi, P., and Pirttila, T., 2005. The role of
C-products in providing customer service – refining the
inventory policy according to customer-specific factors.
International Journal of Production Economics, 93–94,
139–149.Kaminsky, P. and Kaya, O., 2008. Combined make-to-order/
make-to-stock supply chains. IIE Transactions, 41 (2),
103–119.
Kerkkanen, A., 2007. Determining semi-finished products to
be stocked when changing the MTS–MTO policy: case of a
steel mill. International Journal of Production Economics,
108, 111–118.Krupp, J.A.G., 1994. Are ABC codes an obsolete technology?
APICS – The Performance Advantage, 4 (4), 34–35.Li, L., 1992. The role of inventory in delivery-time
competition. Management Science, 38 (2), 182–197.Mennel, R.F., 1961. Early history of the EOQ. APICS
Quarterly Bulletin, 2 (2), 19–22.Olhager, J., 2003. Strategic positioning of the order penetra-
tion point. International Journal of Production Economics,
85 (3), 319.Perona M. and Miragliotta G., 2000. Workload control
techniques: a comparison of theoretical and practical
issues through a survey in field. In: Proceedings of the
11th international working seminar on production economics
– 21–25 February, Igls.Perona M. and Portioli A., 1993. MCG: a new set-up-saving
dispatching rule. In: Proceedings of the IASTED interna-
tional conference on robots and manufacturing, 23–25
September, Oxford, England.Rajagopalan, S., 2002.Make to order ormake to stock: model
and application. Management Science, 48 (2), 241–256.Scarf, H., 1960. The optimality of (S,s) policies in the
dynamic inventory problem. In: K.J. Arrow, S. Karlin and
P. Suppes, eds. Mathematical methods in the social science.
Stanford, CA: Stanford University Press, 196–202.
Silver, E.A., Pyke, D.F., and Peterson, R., 1998. Inventory
Management and Production Planning and Scheduling.
New York: Wiley.Silver, E.A. and Robb, D.J., 2008. Some insights regarding the
optimal reorder period in periodic review inventory systems.
International Journal of Production Economics, 112, 354–366.Soman, C.A., van Donk, D.P., and Gaalman, G.J.C., 2004.
Combined make-to-order and make-to-stock in a food
production system. International Journal of Production
Economics, 90 (2), 223–235.
Soman, C. A., van Donk, D. P., and Gaalman, G. J. C.,
2007. Capacitated planning and scheduling of combined
make-to-order and make-to-stock production in the food
industry: an illustrative case study. International Journal of
Production Economics, 108 (1–2), 191–199.
Spence, A.M. and Porteus, E.L., 1987. Set-up reduction and
increased effective capacity. Management Science, 33 (10),
1291–1301.Stevenson, M. and Silva, C., 2008. Theoretical development
of a workload control methodology: evidence from two
574 M. Perona et al.
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014
case studies. International Journal of Production Research,46 (11), 3107–3131.
Sun, X.Y., et al., 2008. Positioning multiple decouplingpoints in a supply network. International Journal ofProduction Economics, 113, 943–956.
Tsubone, H., Ishikawa, Y., and Yamamoto, H., 2002.
Production planning system for a combination ofmake-to-stock and make-to-order products.International Journal of Production Research, 40 (18),
4835–4851.Van Donk, D. P., 2001. Make to stock or make to order: thedecoupling point in the food processing industries.
International Journal of Production Economics, 69 (3),297–306.
Wilson, R.H., 1934. A scientific routine for stock control.Harvard Business Review, 13, 116–128.
Wu, M.C., Jiang, J.H., and Chang, W.J., 2008. Schedulinga hybrid MTO/MTS semiconductor fab with machine-
dedication features. International Journal of ProductionEconomics, 112 (1), 416–426.
Zsambok, C.E., 1997. Naturalistic decision-making: where
are we now? In: C.E. Zsambok and G. Klein eds.Naturalistic decision-making. Mahwah, NJ: LawrenceErlbaum Associates, 3–16.
Production Planning & Control 575
Dow
nloa
ded
by [
Uni
vers
ity o
f So
uthe
rn Q
ueen
slan
d] a
t 11:
54 0
9 O
ctob
er 2
014