a model-driven dss architecture for delivery management in collaborative supply chains with lack of...
TRANSCRIPT
This article was downloaded by: [141.212.109.170]On: 20 November 2014, At: 07:16Publisher: 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
A model-driven DSS architecture for deliverymanagement in collaborative supply chains with lack ofhomogeneity in productsAndrés Bozaa, M.M.E. Alemanya, F. Alarcóna & Llanos Cuencaa
a Research Centre on Production Management and Engineering (CIGIP), UniversitatPolitècnica de València, Camino de Vera S/N, 46022, Valencia, Spain.Published online: 22 May 2013.
To cite this article: Andrés Boza, M.M.E. Alemany, F. Alarcón & Llanos Cuenca (2014) A model-driven DSS architecture fordelivery management in collaborative supply chains with lack of homogeneity in products, Production Planning & Control:The Management of Operations, 25:8, 650-661, DOI: 10.1080/09537287.2013.798085
To link to this article: http://dx.doi.org/10.1080/09537287.2013.798085
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
A model-driven DSS architecture for delivery management in collaborative supply chains withlack of homogeneity in products
Andrés Boza*, M.M.E. Alemany, F. Alarcón and Llanos Cuenca
Research Centre on Production Management and Engineering (CIGIP), Universitat Politècnica de València,Camino de Vera S/N, 46022 Valencia, Spain
(Received 12 April 2013; final version received 12 April 2013)
Uniform product deliveries are required in the ceramic, horticulture and leather sectors because customers requireproduct homogeneity to use, present or consume them together. Some industries cannot prevent the lack of homoge-neity in products in their manufacturing processes; hence, they cannot avoid non-uniform finished products arrivingat their warehouses and, consequently, fragmentation of their stocks. Therefore, final uniform product amounts do notmatch planned production ones, which frequently makes serving previous committed orders with homogeneous quan-tities impossible. This paper proposes a model-driven decision support system (DSS) to help the person in charge ofdelivery management to reallocate the available real inventory to orders to satisfy homogenous customer requirementsin a collaborative supply chain (SC). The DSS has been validated in a ceramic tile collaborative SC.
Keywords: decision support system; lack of homogeneity in products; delivery management; collaborative supply chain
1. Introduction
Lack of homogeneity in products (LHP) appears in thoseindustries whose productive processes cannot eliminatethe heterogeneity of their inputs (i.e. raw materialobtained from nature). LHP also appears in the produc-tive processes that introduce some variability into theiroutputs due to uncontrolled productive factors, evenwhen their inputs are homogeneous (Alemany et al.2013). LHP may become a considerable problem whencustomers acquire several units of a given product andrequire product homogeneity to use, present, arrange orconsume them together in order to avoid functional oraesthetical problems. A slight difference in a part is eas-ily seen in products such as parquet strips, leatherwear,floor tiles or pearl necklaces (Alarcón et al. 2011).
Companies with LHP cannot foresee the final charac-teristics of those products in lots defined by the produc-tion plan because they cannot prevent non-uniformfinished products arriving at their warehouses and, conse-quently, fragmentation of their inventories. These compa-nies are obliged to include one or several classificationstages along the production process. In the classificationstage, units of the same product are sorted into homoge-neous subgroups (subtypes) based on certain attributes(classification criteria) that depend on each sector, whereeach subtype includes uniform final products. Hence,the products included in each subtype meet customer
expectations in terms of functional and/or aestheticshomogeneity. In LHP contexts, customer orders manage-ment becomes more complex and, therefore, the orderpromising (OP) process plays an important role in thistask. The OP process includes the set of business activi-ties that are triggered to provide a response to customerorder requests (quantity, delivery date, etc.). In this pro-cess, is necessary to compute if there are enough real orplanned products available that have not been previouslycommitted (ATP – available to promise).
Traditional ATP assignment systems assume producthomogeneity; therefore, they allow the accumulation ofcurrent uncommitted stocks from warehouses (real ATP)and from planned amounts (planned ATP) of differentmanufacturing lots and periods. However, companieswith LHP obtain non-uniform lots and the real homoge-neous quantities remain unknown until production hasfinished. This aspect becomes a problem when customerorders should be promised based on the planned ATPquantities where real homogenous quantities are notknown in advance. This means that, more often than not,some ATP-based orders cannot be fulfilled once theproduction of lots has finished because the amount ofavailable homogeneous product does not suffice. Thissituation causes problems in committed orders withimmediate delivery dates. For such cases, by reallocating
*Corresponding author. Email: [email protected]
Production Planning & Control, 2014Vol. 25, No. 8, 650–661, http://dx.doi.org/10.1080/09537287.2013.798085
� 2013 Taylor & Francis
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
the available inventory of each subtype in warehouses(real inventory) among all the orders, it is possible toprioritise those orders with immediate delivery dates.
The decision to reallocate current stock to customerorders becomes complicated in those companies with lotsof orders that have different delivery dates and severalproducts in each order (order lines). Furthermore, thissituation becomes even more complicated in the LHPcontext, given the fragmentation of inventory becauseseveral subtypes of the same product exist and thehomogeneity requirements imposed by the customer.This means having to introduce new constraints whenreallocating stock to orders.
Therefore, LHP increases not only the volume ofinformation to be used, but also the number of possiblecombinations when serving orders, and all of them withdifferent profits. All these factors complicate the task ofobtaining solutions that are not just optimum, but alsofeasible. These additional complexity aspects for reallo-cating stock justify the use of tools to help withdecision-making, i.e. mathematical programming models.The solutions obtained with such tools may help thesupply chain (SC) minimise the impact of LHP on thecustomer service level. Hence, for the purpose of provid-ing support to delivery management tasks, a model-dri-ven decision support system (DSS) to optimallyreallocate the inventory of homogeneous subtypes toLHP-characterised promised orders in collaborativemake-to-stock (MTS) environments is proposed. Themodel solution provides orders that can be served withreal inventory and from the subtype of which uniformproduct can be delivered, following the maximisation ofprofits and orders completed with earlier delivery dates.
The rest of the paper is outlined as follows: Section 2includes a review of the relevant literature relating to thepaper. Section 3 describes the developed model-drivenDSS for delivery management in collaborative SCs withLHP. Section 4 presents the DSS application to a collab-orative ceramic tile SC case. Finally, Section 5 offers themain conclusions and future research lines.
2. Literature review
The creation of SCs has offered firms new businessopportunities based on collaboration and coordination.Multiple benefits have been identified, thanks to the col-laboration and coordination in SCs (Bititci et al. 2007).
For collaboration to be effective, it is necessary toadapt the distinct SC processes; for instance, the demand-forecasting process (Poler et al. 2008), the demandmanagement process (Abid, D’Amours, and Montreuil2004) or the reverse logistic (Hernandez et al. 2011). Inthe demand management domain, the OP process (Abid,D’Amours, and Montreuil 2004) and the ATP allocationare extremely important matters.
In the short term, the customer orders previouslycommitted by ATP allocation should be completed fordelivery in order to meet the promised due date. Yet,owing to unforeseen events, there may not be enoughavailable stock in the right quantities to cover theseorders and to meet the promised due dates. In order tofind a satisfactory solution for both customers and SC,some shortage planning models (Pibernik 2006; Zschorn2006) have been developed. Indeed, shortage planningdeals with the activities to be accomplished, should stocknot be available and includes certain decisions likecustomer negotiation (Framiñán and Leisten 2009),outsourcing (Bhakoo, Singh, and Soha 2012) or substitu-tive products (Balakrishnan and Geunes 2000).
Nonetheless, despite collaboration being important inthe order management, very few works have addressedthis issue (Alarcón, Alemany, and Ortiz 2009; Alemanyet al. 2008) perhaps due to the added complexity thatcoordination and collaboration entail. This complexitymay be further complicated if, besides, the OP processmust consider LHP problems (Alarcón et al. 2011). Insuch cases, the vast quantity of up-to-date information tobe processed, and in real time, plus the swift, reliablerequirements involved in replying to customers, renderthe use of decision-making tools, like DSSs, virtuallycompulsory.
DSSs can be developed so they can utilise knowl-edge and handle knowledge sources as effectively aspossible (Kubat, Öztemel, and Taşkιn 2007). One impor-tant advantage of DSSs is that the decision-maker neednot understand the complexities of mathematical model-ling (Gomes da Silva et al. 2006). The capacity of DSSsto be able to dynamically respond to changes and tocommunicate with the affected SC nodes could also bestressed and is a basic requirement for SC collaborationto work (Jagdev and Thoben 2001), and the technologi-cal advances in the Internet make available thesecollaborative DSSs (Boza, Ortiz, and Cuenca 2010).
Although relevant research has been done on DSSsin relation to topics such as order planning (Azevedoand Sousa 2000), order management (Abid, D’Amours,and Montreuil 2004) and shortage planning (Okongwuet al. 2012), to our knowledge, there are no works thatpropose mathematical programming models either inisolation or within the DSS framework which addressreallocating inventories for shortage planning in collabo-rative contexts with LHP. Thus, there is a gap in theliterature which justifies the proposal set out in the nextsection.
3. Model-driven DSS architecture for collaborativedelivery management in SCs with LHP
This section not only describes the context and problemsinvolved in delivery management in SCs with LHP, but
Production Planning & Control 651
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
also proposes their management within a collaborativeframework. Then, it identifies LHP stock reallocation asa solution for shortage situations when delivering ordersand proposes a mathematical programming modelprecisely for this purpose integrated into a DSS.
3.1. Context: towards a new collaborative framework
The physical SC configuration is assumed to be com-posed of three kinds of nodes: several production plants;one central warehouse; and several selling points (SPs)(Figure 1). On the one hand, production plants manufac-ture in lots following an MTS strategy. All the productsmay be processed in any production plant and, owing toLHP, each manufactured lot of a given product may giverise to different product subtypes of this product. Onceclassified, quantities of the same subtype are placed inthe central warehouse with other equal subtypes.
Customers place orders at SPs, while orders to bedispatched and delivered to each SP are prepared in thecentral warehouse. There are two kinds of SPs (Figure 1):those belonging to own enterprise selling networks (ownselling points – OSPs) and other SPs that belong to theindependent selling network (independent selling points– ISPs). These ISPs also commercialise other productsfrom other SCs.
The initial relation of ISPs with the rest of the SC islimited to orders delivery. In order to avoid problemswith customers owing to delays or failures in supply,ISPs tend to outsize orders and request an earlier deliv-ery date to the one agreed on with the customer. Indeed,one usual practice consists in grouping orders, whichentails having to deliver larger homogeneous quantitiesthan those each customer actually ordered at ISPs. Thismakes inventory allocation a complex issue because if
the LHP is taken into account, then the larger theamount of homogeneous units required, the more diffi-cult it is to serve the order. These ISPs practices generatereserves of ‘fictitious’ inventory and leave a largeramount of product unavailable for OSPs which areobliged to delay due dates, which worsens customer ser-vice. These practices involve significant drawbacks forthe SC as a whole.
In order to improve this situation, greater collabora-tion of ISPs and the rest of the SC is as considerednecessary. That is, moving towards a collaborative SCwhere entities agree on a set of commonly definedobjectives and use their complementary assets to gainlong-term competitive advantage (Lejeune and Yakova2005). Along these lines, it is worth bearing in mind thedeterminant role of collaboration and coordination inimproving the SC’s logistic processes; for instance, inLambert and Cooper (2000), Romano (2003), Holwegand Pil (2008), or in the many works cited in the reviewby Arshinder, Kanda, and Deshmukh (2008).
According to (Akkermans, Bogerd, and van Dorema-len 2004), it is important that SC members share infor-mation, basically on current orders, the productionstatus, plans and forecasts. In line with this, the newproposed collaborative framework is based on the ISPs’increased trust in the SC and the improved informationflow which allows the central warehouse to obtain realand accurate knowledge on the orders (quantities anddue dates) managed at the ISPs. Within this new collabo-rative framework, ISPs share information and the centralwarehouse can centralise decision-making about theallocation and reallocation of stocks to orders comingfrom the OSPs and ISPs, which may be considered themaximum degree of collaboration (Alemany et al. 2011)to help achieve the highest level of customer service.
Main RawMaterials
Suppliers (fritsand enamels)
Others RawMaterialsSuppliers
ManufacturingPlants
OtherManufacturing
Plants
Raw MaterialsSuppliers
CentralWarehouse
Other CentralWarehouses
Own Selling Points
IndependentSelling Points
SUPPLY CHAIN
Collaboration Framework
Figure 1. The physical SC configuration and the DSS collaboration framework.
652 A. Boza et al.
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
In this way, and if required, with the new DSSproposed, it is possible to use the products reserved byISPs to cover more urgent orders from OSPs and viceversa. This has important managerial implications; thisnew framework of relations based on trust in coordina-tion and collaboration permits ISPs to be completelyintegrated into the SC as a whole.
The following section describes the mixed integer lin-ear programming model proposed for stock reallocationwhich is used by the DSS, should there be shortages.
3.2. Mathematical programming model for LHP stockreallocation (SR-LHP)
This section presents the main characteristics, assumptionsand limitations of the LHP stock reallocation problem.Then, the mathematical programming model proposed bythe authors to solve it is described. The components of themathematical programming model are introduced alongwith the problem presentation to facilitate the later modelformulation understanding. The nomenclature used for theSR-LHP-1 model appears in Figure 2.
As planned production lots are manufactured andreceived in the central warehouse, they arrive classifiedinto the corresponding subtypes. At this time, checks aremade to see if there is a sufficient amount of the uniformsubtypes obtained to serve the promised orders. Owingto LHP, it is quite usual that all orders with immediatedelivery dates cannot be completely served with thereserved stock because the obtained amount of the differ-ent quantities of uniform products differs from thoseplanned, so there is not enough to complete all them.Due to manufacturing lead times, it is not possible toobtain new production lots on time in order to completethese orders. An alternative to avoid or minimise delaysdue to LHP as much as possible consists in reallocatingthe current inventory of each subtype (b) of each product(k) in the warehouse (qkb) from orders with no immedi-ate delivery dates to those with immediate delivery dates.When reallocating the LHP current stock and preparingorders for delivery, the following assumptions and limita-tions should be considered:
Indices Setsi each order with a delivery date within the horizon
(i=1...I) l each order line that makes up the orders being
considered (l=1...L) k each existing product in considered orders (k=1...K) b each existing product subtype in considered products
(b=1...B)
I(he) the set of orders whose delivery date is within the delivery horizon
L(i,k) the set of order lines l, which contain product k and correspond to order i.
B(k) the set of existing subtypes of product k K(i) the set of products k which are included in
order iParameters
p1 A specific weight that the decision-maker confers to the profit in the objective functionp2 A specific weight that the decision maker confers in order to serve orders with earlier delivery dates in the
objective functionsi Profits from order i h Reallocation horizon. This is the delivery deadline date which determines the orders to be considered by the
model; orders whose delivery dates are equal to or before the horizonOrders delivery horizon. This date determines which orders should be prepared in the warehouse to be
subsequently dispatched and delivered to the customer: orders whose dispatch dates are equal to or before the delivery horizon
fdi Dispatch date of order i from the central warehousenI(he) Number of orders whose dispatch date are within the delivery horizon ( ).
nli Number of order lines that order i hasdkli The ordered amount of product k, in line l of order i qkb The amount of product k and subtype b available in the warehouse
smax Maximum of .
smin Minimum of .
ε A positive and lower value than the unit Decision variables
Yi Binary variable that takes a value of 1 if order i is completely reserved (all its L(i) lines are reserved with the real inventory), and a value of 0 otherwise
Uklib Binary variable that takes a value of 1 if line l of order i has a current inventory reserve of product k and subtype b and a value of 0 otherwise
ATP0kb The real inventory amount of product k and subtype b, which is not reserved after having reassigned the current inventory and, therefore, it remains available to compromise with other orders.
eh
eh
is
is
Figure 2. Nomenclature.
Production Planning & Control 653
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
• Each customer order i is composed of severalorder lines l. For each order line l, the quantityof a specific product k in order i (dkli) is known.All the lines of the same order have the samedue date, implying the time bucket when thecustomer requests goods to be received. Due totransportation times, the order may have to bedelivered before the due date, thus rendering itnecessary to differentiate between due dates anddelivery dates ( fdi).
• Orders with immediate delivery dates are thosewith a delivery date ( fdi) lower than the deliveryhorizon (he), i.e. fdi6 he. The delivery horizon(he) represents a period length immediately afterthe current point of time needed to prepareorders to be immediately delivered. This concepthelps identify the orders that should be preparedimmediately for delivery (I(he)) in order to meetpromised customer due dates, and shouldconsequently be completed with the current LHPstock because there is no time to launch newproduction lots.
• The overall customer orders considered whenreallocating the current LHP stock are thecommitted orders with a delivery date within thedelivery horizon (fdi6 he) and other orders withsome order lines reserved with the current LHPstock with a delivery date within the reallocationhorizon (h).
• Partial deliveries are not allowed; that is, all theorder lines of an order are jointly served.
• The requested quantity of a certain product in anorder line (dkli) should be served throughhomogeneous units of this product. Hence, eachorder line is completed with a unique subtype(b) and it is not possible to mix differentsubtypes to serve an order line in order to guar-antee homogeneity.
Given the huge volume of orders, solving the abovereallocation problem becomes a very difficult and time-consuming task. In this context, mathematical program-ming models have been proved to be useful tools. In thispaper, the authors propose a mixed integer programmingmodel to support the decision-making process of reallo-cating LHP stock to orders with immediate delivery,dubbed Stock Reallocation-LHP-1 (SR-LHP-1), as thecore of a more sophisticated DSS. The SR-LHP-1 modelreallocates available LHP stock among all the inputorders by following two objectives: (1) maximise theprofits of served orders with the LHP stock and (2)maximise the number of orders served with early deliv-ery dates. The SR-LHP-1 model output provides thedecision-maker with already committed orders i thatmust be served from the LHP stock (Yi= 1); for all their
order lines l, the specific subtype b of product k fromthe current LHP inventory is used (Uklib = 1). TheSR-LHP-1 model also computes each subtype’s remain-ing current inventory for all the products that has notbeen reserved by any order after reallocation; thus, itbecomes ATP other customer orders (ATP0kb).
The model formulation proposed to reallocate thecurrent inventory to immediate delivery orders is asfollows:
3.2.1. Objective function
As previously noted, the objective function includes twoobjectives. The first objective (1) aims to maximise thetotal profits obtained from reallocations because onlythose orders finally reserved with the current LHP stock(Yi= 1) contribute to the total profits.
Max½z1� ¼Xi
si � Yi (1)
The second objective (2) attempts to maximise thenumber of orders reserved from the current inventorywith earlier delivery dates, which is the equivalent ofmaximising the total sum of the difference between thereallocation horizon and the delivery date, because if h isa fixed quantity, then the lower the fdi, the greater theh-fdi difference. Based on this objective, priority is givento those orders with earlier delivery dates despite theirassociated profits because it is assumed that for thosewith later delivery dates, there is more flexibility to findother solutions to serve them (i.e. modifying the masterplan with additional production lots). In order to avoidthe SR-LHP-1 model not serving those orders with maxi-mum delivery dates (fdi= h), and as their contribution tothe maximisation of z2 is zero (h-fdi= 0), parameter ɛ (apositive value lower than the unit) is used to induce themodel to serve as many orders as possible. Throughparameter ɛ, the above orders also contribute to themaximisation of z2. Therefore, if there is enough LHPstock, the solution first chooses to serve them (Yi= 1).
Max½z2� ¼Xi
(h� fdi þ e) � Yi (2)
To avoid allocating current LHP stock to orders withvery high associated profits, but also very long deliverydates before other orders with lower associated profits;but, with earlier delivery dates, the simultaneous consid-eration of both objectives is proposed. To that end, theyare placed together in a single objective by the completeaggregation procedure, which consists in the sum of thetwo objectives according to weights p1 and p2. Theseweights are assigned by the decision-maker in sucha way that the heavier the weight, the greater the
654 A. Boza et al.
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
importance for the decision-maker. As both profits anddelivery dates are terms of a very different significance,it is necessary to grade them so that both objectives fallwithin a comparable range which, in this case, is [0, 1].Finally, the overall objective function appears as in (3).
Max½z� ¼ p1Xi
si � smin þ esmax � smin
� �� Yiþ
p2Xi
h� fdi þ eh
� �� Yi ð3Þ
Constraints:
Xi2I(he)
Yi ¼ nI(he) (4)
nli �Xk2K(i)
Xl2L(i)
Xb2B(k)
Uklib
!� (1� Yi) � nli 8i (5)
Uklib � Yi 8i; k 2 K(i); l 2 L(i; k); b 2 B(k) (6)
Xi
Xl2L(i)
dkil � Uklib þ ATP0kb ¼ qkb 8k; b 2 B(k) (7)
Xb2B(k)
Uklib � 1 8i; k 2 K(i); l 2 L(i; k) (8)
Uklib; Yi 2 f0; 1g 8i; k 2 K(i); l 2 L(i; k); b 2 B(k) (9)
ATP0kb � 0 8k; b (10)
Constraint (4) forces orders with delivery dateswithin the delivery horizon to be served. That is, abso-lute priority is given to reserve the current available LHPstock for these orders.
Constraints (5) ensure that the orders with a numberof reserved lines below its number of order lines are notserved (Yi= 0). Through these constraints, the orderswhose lines have all been allocated can or cannot beserved (Yi= 0,1). Therefore, additional constraints (6) arenecessary to force orders whose order lines have all beencompleted with the current LHP inventory to be served(Yi= 1). Only these orders contribute to the objectivefunction. Furthermore, Constraints (6) act in such a waythat should an order not be served (Yi= 0), none of itsorder lines are reserved with the LHP stock.
Constraints (7) represent balance equations by ensur-ing that the amount of each subtype b of product k allo-cated to the different order lines, plus the amount notallocated (uncommitted: ATP0kb), must equal the amountof that product subtype available in the warehouse. As
variable ATP0kb is non-negative, the allocated LHP stockcan never exceed the initial available quantity (qkb) inthe warehouse.
Constraints (8) force each single order line to becompleted only with a product subtype. This constraintensures the required product uniformity in LHP environ-ments, and along with considering several subtypes ofeach product, it allows LHP characteristics to be prop-erly modelled.
Constraints (9) define decision variables Uklib and Yias binary variables.
Finally, Constraints (10) indicate that the remainingcurrent stock of product k and subtype b not allocated toany order is a continuous variable that should benon-negative (ATP0kb).
Owing to Constraint (4), the model solution may notbe feasible if there is not enough uniform products avail-able in the warehouse to cover all the order requirementswith immediate delivery dates. In this case, it is possibleto remove Constraint (4) from the model SR-LHP-1 andto solve the resulting model again, dubbed as SR-LHP-2.This new SR-LHP-2 model always provides thedecision-maker with a feasible solution that also givespriority to orders with earlier delivery dates.
3.3. DSS architecture
To facilitate the utilisation of the above models bymanagers and to provide additional functionalities, amodel-driven DSS is proposed (SR-LHP-DSS). In thissection, the SR-LHP-DSS architecture for the deliverymanagement of LHP SCs is described in accordancewith the modules presented in Figure 3 (extraction, pre-processing, solver, post-processing and analysis).
3.3.1. Extraction
The input data for SR-LHP models (parameters, indicesand sets) originate from different information systemsbelonging to the company and the SPs. In the proposedSR-LHP-DSS, information about customer orders isextracted from sales information systems. In the new col-laborative framework, the ISPs provide the DSS with thereal orders placed by customers (real quantities and duedates) instead of consolidating all the orders, enlargingthem and requesting an earlier delivery date. For SCswith LHP, it is very important to know the exact ordersize to facilitate the achievement of the homogeneityrequirement. Sharing information among collaborativeSC members is essential for better decision-making. Thisinformation includes: the delivery date of each order, theprofits obtained and the amount of product requested foreach order line. Information about the amounts of prod-ucts and their subtypes is obtained from the informationsystems of each production plant and the transport timeis obtained from the logistics information system.
Production Planning & Control 655
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
3.3.2. Pre-processing
Before running the model, it is necessary to performprevious data processing. This involves calculating thenumber of order lines in each order (nli), the number oforders whose delivery date is within the delivery horizon(nI(he)), the delivery date of each order from the centralwarehouse ( fdi) computed as the difference betweenthe order due date ( fei) and transport time (tsp) fromthe central warehouse to the SP ( fdi = fei�tsp), the maxi-mum (smax) and minimum (smin) profit of the ordersprocessed to scale profit and, finally, the total amount ofsubtypes of products as the sum of the amounts ofsubtype (b) of products (k) in the production plants (p):qkb ¼
Ppqpkbp.
3.3.3. Solver
Before reaching the solving stage, the decision-makershould introduce the weights allocated to each objectiveconsidered. Firstly, the SR-LHP-1 model is solved: if asolution is obtained, it means that all the orders in thedelivery horizon can be served. If an unfeasible solutionarises, model SR-LHP-2 is run, which always provides afeasible solution but, in contrast, some orders are notserved. The solution obtained includes information aboutif one order will be completely reserved or not, the sub-type of product be reserved for each order line and theamount of the physical stock of each subtype of productbe unreserved. The decision-maker may confer differentvalues to the weights of each objective and can run thecorresponding model several times generating a ‘what-if’scenario.
3.3.4. Post-processing
Post-processing is proposed to provide additional infor-mation by comparing the different solutions obtainedfrom modifying the weights allocated to each objective.The SR-LHP-DSS calculates specific performanceparameters for each solution (Figure 4), which allow thedecision-maker to compare them.
3.3.5. Analysis
The analysis stage must enable the decision-maker toconsult the results of the model in as much detail as pos-sible for each solution generated. Through this module,the decision-maker may analyse how sensitive the solu-tion is to these weights and to compare them with othersolutions.
To this end, having executed the model and havingobtained the performance parameter values, the distanceof these parameters to the best value of the performanceparameters obtained in the solutions to that time is mea-sured and saved from each parameter in the post-treat-ment stage. This stage involves calculating not only thebest value of all the values saved for each parameteruntil that time (Figure 5), but also deviation in terms ofthe best values for each solution and for each parameterof the solution (Figure 6).
3.3.6. Decision
As explained, during the stock reallocation process, theinformation shared among collaborative SC members isused to obtain a solution. Therefore, from a collaborative
Figure 3. The SR-LHP-DSS architecture.
656 A. Boza et al.
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
perspective, this aspect facilitates better decision-making.However, it is possible to execute the decision process,should unforeseen events occur (event-driven), i.e. arrivalof urgent orders. Finally, based on the analysis of theresults, the decision-maker chooses a solution from theexisting ones to be implemented.
4. SR-LHP-DSS validation: case of a ceramic tile SC
The model-driven SR-LHP-DSS is validated by develop-ing a prototype for a real ceramic tile SC. The SR-LHP-DSS results are compared with the manual stockreallocation process carried out by the company. Forconfidentiality reasons, a factor is applied to the overallprofits in both cases.
The ceramic tile SC includes three production plants,a central warehouse and 28 SPs, of which 11 are OSPand 17 are ISP. The LHP in the ceramic SC brings aboutchanges in the tone and gage of ceramic tiles. As cus-tomers require the same tone and gage in their orders,product subtypes (combination of product-tone-gage) areidentified, stored and managed separately in the centralwarehouse. Besides, orders are prepared in this centralwarehouse to be dispatched and delivered to the SPs(OSP and ISP), where final customers place their orders.
Currently, the ATP policy is First Come, First Com-mitted, i.e. orders are committed as they arrive. Due tothis ATP commitment policy and the own LHP effects,
Figure 4. The parameters obtained with the post-treatment.
Figure 5. Calculation of the best values of each parameter inthe post-treatment of all the saved solutions j.
Figure 6. Calculation of the deviations for each parameter andsolution in relation to the best values of each parameter in thepost-treatment.
Production Planning & Control 657
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
the person in charge of deliveries usually needs to reallo-cate inventories, should stock shortage take place tocomplete orders for immediate delivery. Although theperson in charge of managing deliveries has a computertool that quickly locates product lines for which there isnot enough stock (incomplete lines), finding a new solu-tion to complete these lines through the manual realloca-tion of current stock is very time-consuming, and it iseven possible that this solution does not exist.
Data on the manual stock reallocation carried out bythe ceramic SC are captured to compare them with theSR-LHP-DSS solutions. The considered reallocationhorizon is one year. This manual process is performedwith 2274 orders in the planning horizon, with 9389lines and an average of 4.12 lines per order, a maximumof 108 lines and a minimum of one line per order. Theprofit made with these orders is e8,648,346.05. How-ever, there are 934 orders in the order book within thedelivery horizon (two weeks) with a total profit ofe4,808,518.84, and the manual reallocation completes787 of these orders, implying a profit of e3,237,986.38.
The SR-LHP-DSS helps managers complete thoseorders with immediate delivered, thus fulfilling orderdelivery dates in the collaborative SC. The technologiesused to construct the SR-LHP-DSS prototype are: Java
v7 and the ECLIPSE platform to develop dialogcomponents; MPL 4.2 and the solver CPLEX to translatethe mathematical programming models to a readablemachine format; and a Microsoft Access database tostore the corresponding data.
The extraction and pre-processing processes arecarried out to obtain the input from the ceramic tileinformation systems in order to provide the SR-LHPmodels with the same data used in the manual process.Next, the solver is executed to evaluate these data inthree solution scenarios generated through differentobjective function weight definitions (Figure 7).
With these input data, the SR-LHP-1 model solutionproves unfeasible for the all scenarios because there isnot enough homogeneous stock to serve all the orderlines in the horizon; so, it is not possible to fulfil theConstraint [4] of this model. Hence, the SR-LHP-2model is executed and optimal solutions are found. Forthis second model, the computational time of 4.62 s spentto find the optimal solution of W_B case shows themodel’s utility for real size problems.
The interface design for capturing the performanceparameters of the solutions obtained by the SR-LHP DSSfor the various scenarios in the post-treatment phase areshown in the following screen (Figure 8). As illustratedin Figure 8, we find: the W_DD scenario, which producesoptimal with early delivery dates, completes 1857 orderswith a profit of e5,338,811.74; the W_B scenario obtainsthe same number of completed orders, but these ordersprovide more profit (e5,438,283.74); and finally, theW_P scenario completes 1774 orders (4.47% less than
Solution Name Weights
Profits (p1) Delivery Date (p2) (W-DD) Weight in delivery dates 0 1 (W-B) Weights balanced 0.5 0.5 (W-P) Weight in profits 1 0
Figure 7. Scenarios definition for the SR-LHP DSS.
Figure 8. The performance parameter values for W_DD, W_B and W_P.
Orders Completed inside h e
Manual & DSS Difference
Profits Manual & DSS
Difference
Company Manual 787 --- 3,237,986.38 ---
W_DD 832 5.72% 3,539,720.29 9.32%W_B 824 4.7% 3,570,618.21 10.27%W_P 766 -2.67% 3,628,407.22 12.06%
DSS Scenarios
Figure 9. The manual and SR-LHP-DSS reallocations inside he.
658 A. Boza et al.
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
the W_DD scenario), but its profit is e5,546,099.32(3.88% more than the W_DD scenario). Thus, the W_Pscenario completes fewer orders than the W_DD sce-nario, but it chooses orders that provide a higher profit.After analysing these findings, as expected, the weightsassigned to each objective affect the obtained results, thusvalidating the proposed SR-LHP models. Furthermore,the possibility for the decision-maker to generate multiplesolutions by changing the objective weights provides a‘what-if’ scenario, which has proved to be one of themain functionalities for each DSS.
The decision-maker may also be interested in obtain-ing more details of the results for the delivery horizonbecause he/she requires an immediate decision (profits,orders completed and uncompleted in he in Figure 8).The validation allows the comparison of the DSS resultsobtained by the manual reallocation process carried outin the delivery horizon by the company. Figure 9 offersthis comparison using the different objectives weights.
As observed in Figure 9, manual reallocationcompletes 787 of the 934 customer orders in the deliveryhorizon, with a profit of e3,237,986.38. However, theSR-LHP-DSS reallocation obtains more profit than themanual one for all the scenarios. The W-DD scenarioincludes the largest number of completed orders (832),that is 5.72% more orders than the manual reallocation,as well as 9.32% more profit than the manual realloca-tion. The W-B scenario completes fewer orders than theprevious scenario (824), but obtains more associated prof-its (10.27% more than the manual reallocation). The W-Pscenario completes the smallest number of orders (766),but makes the most profits (12.06% more than the manualreallocation). The superiority of the SR-LHP-DSS resultsin all the generated scenarios as compared to the manualreallocation procedure, plus the reduced solution time ofthe SR-LHP models for real cases, prove the validity ofDSS and its convenience in real problems.
Finally, based on the DSS analysis functionality, thedecision-maker must decide on satisfying more custom-ers (more completed orders) – the solutions have manyorders not completed – or gaining more benefits not onlyduring the current time period (he), but also possibly inthe future (h).
5. Conclusions and future research
Due to LHP, differences between planned homogeneousquantities and the finally manufactured ones are frequent,which made it impossible to serve all promised orderson time with the initial ATP allocation. This paperdescribes a model-driven DSS to overcome shortagesituations during the delivery management of customerorders for collaborative SCs with an MTS strategy char-acterised by LHP. The DSS core is composed by twonovel mathematical programming models with multiple
weighted objectives to support LHP-stock reallocations.To facilitate mathematical models utilisation by manag-ers, different functionalities are included in the DSS.Along these lines, and before making a definitivedecision, the decision-maker can perform different ‘what-if’ simulations (scenarios) by modifying the weightallocated to each objective, and by assessing andcomparing different solutions based on diverse perfor-mance parameters.
This model-driven DSS has been validated in acollaborative ceramic SC, where LHP materialises in theform of different tones and gages in ceramic tiles. Thestock reallocations obtained from the DSS in variousscenarios have been compared with the ceramic tile com-pany’s current procedure. The results demonstrate thatthe DSS outperforms the manual procedure employed bythe ceramic company for all the scenarios generated.
Therefore, the main implications for managersdeduced from its use are the following:
(1) Considerable time saving in solving the realloca-tion problem and better quality of solutionsobtained; the DSS immediately allows knowing ifthere is enough homogeneous inventory availableto deliver all the orders within the delivery horizonand, if so, it provides the optimal solution. Whenno feasible solution exists, the DSS allows deter-mining the orders to be allocated with the homoge-neous inventory to obtain an optimal solution.
(2) Support for generating different scenarios andoptimal solutions in a user-friendly way, bychanging the weights of the objectives, the DSSgenerates as many solutions as the decision-makerdesires.
(3) Support for analysing and comparing the differentsolutions in order to choose the most satisfactoryone.
Furthermore, the DSS also favours the developmentof a new collaborative context based on greater trustthrough information sharing and a centralised decision-making. In this new context, ISPs no longer transferinformation on consolidated orders, which are ofteninflated, to the central warehouse, but information abouttheir customers’ real orders. This information is incorpo-rated into the DSS for supporting the centraliseddecision-making process of reallocating available homo-geneous stocks to orders coming from both, the OSPsand ISPs, when shortage situations occurs.
From a practical point of view, the real customers’orders are much smaller than those consolidated by theISPs that make easier to find homogeneous units tocomplete them. Because shortage situations are frequentdue to LHP, better reallocation of available stock is madeto serve more priority orders. All these aspects willimprove SC efficiency by reducing SC inventory holding
Production Planning & Control 659
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
costs, stocks fragmentation and increasing the customerservice level.
Future research lines should consider other objectivesderived by applying the SR-LHP-DSS to other sectors.Furthermore, a goal programming model that providesthe decision-maker with non-dominated solutions shouldbe included in the SR-LHP-DSS for future implementa-tions. Finally, in order to minimise future stock realloca-tion due to LHP as much as possible, a knowledgedatabase that collects past patterns of homogeneousquantities in production lots can be designed to predicthomogeneous outcomes. This information can be used inproduction planning to define lot sizes and can be usedlater during the OP process as a proactive action to LHP.
AcknowledgementsThis research has been carried out within the framework of theproject funded by the Spanish Ministry of Economy andCompetitiveness (Ref. DPI2011-23597) and the PolytechnicUniversity of Valencia (Ref. PAID-06-11/1840) entitled‘Methods and models for operations planning and ordermanagement in supply chains characterized by uncertainty inproduction due to the lack of product uniformity’ (PLANGES-FHP). Also, we thank the comments and suggestions made bythe Editors and the Reviewers. In our opinion, these changeshave improved the quality of the paper.
Notes on contributorsAndrés Boza is an assistant professor inManagement Information Systems at theUniversitat Politècnica de València (UPV)in Spain. He is a computer science engineerand received his PhD in SC Managementand Enterprise Integration from the UPV.He is member of Research Centre onProduction Management and Engineering(CIGIP) where he has participated in the
doctoral program and in national and international projectsabout information systems for production management, SCmanagement and performance measurement. He has publishedseveral papers in books, journals and international conferencesin these fields.
M.M.E. Alemany is an associate professorin Operations Management and OperationsResearch at the Universitat Politècnica deValència (UPV) in Spain. She is anindustrial engineer and received her PhD inIndustrial Engineering from the UPV. She isa member of Research Centre onProduction Management and Engineering(CIGIP) and has participated in several
Spanish and European Projects. Currently, she is leading aSpanish Government Project. Her research is focused onHierarchical Production Planning, Collaborative OP Processand SC Management under conceptual/quantitative perspective.She has several research papers in books, conferences andinternational journals.
F. Alarcón is an associate professor inOperations Management and OperationsResearch at the Universitat Politècnica deValència (UPV) in Spain. He is anindustrial engineer and received his PhD inIndustrial Engineering from the UPV. He isa member of Research Centre onProduction Management and Engineering(CIGIP) and has participated in different
Spanish and European Projects. His research is focused onEnterprise Modelling and Integration, BPM, Collaborative OP,Collaborative Operations Planning in Supply and DistributionNetworks. He has published in international journals andconference proceedings.
Llanos Cuenca is an assistant professor inEnterprise Management and ManagementInformation Systems at the UniversitatPolitècnica de València (UPV) in Spain.She is a computer science engineer andreceived her PhD from the UPV. She is amember of Research Centre on ProductionManagement and Engineering (CIGIP). Shehas participated as a researcher on several
Spanish and European projects. At this moment, her researchlines focus on Interoperability, Enterprise Architectures,Strategic Alignment and the use of ICT in these areas. Asresult of this research, she has published several papers ininternational conferences and journals.
References
Abid, C., S. D’Amours, and B. Montreuil. 2004. “CollaborativeOrder Management in Distributed Manufacturing.” Interna-tional Journal of Production Research 42 (2): 283–302.
Akkermans, H., P. Bogerd, and J. van Doremalen. 2004.“Travail, Transparency and Trust: A Case Study ofComputer-Supported Collaborative Supply Chain Planningin High-Tech Electronics.” European Journal of Opera-tional Research 153 (2): 445–456.
Alarcón, F., M. M. Alemany, F. C. Lario, and R. F. Oltra.2011. “La Falta De Homogeneidad Del Producto (FHP) EnLas Empresas cerámicas Y Su Impacto En La reasignaciónDe Inventario [The lack of homogeneity in the product(LHP) in the ceramic tile industry and its impact on thereallocation of inventories].” Boletín De La SociedadEspañola De Cerámica Y Vidrio 50 (1): 49–58.
Alarcón, F., M. M. E. Alemany, and A. Ortiz. 2009. “ConceptualFramework for the Characterization of the Order PromisingProcess in a Collaborative Selling Network Context.” Inter-national Journal of Production Economics 120 (1): 100–114.
Alemany, M. M. E., F. Alarcón, F. Lario, and J. J. Boj. 2011.“An Application to Support the Temporal and Spatial Dis-tributed Decision-Making Process in Supply Chain Collab-orative Planning.” Computers in Industry 62 (5): 519–540.
Alemany, M. M. E., F. Alarcón, A. Ortiz, and F. C. Lario. 2008.“Order Promising Process for Extended Collaborative SellingChain.” Production Planning & Control 19 (2): 105–131.
660 A. Boza et al.
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14
Alemany, M. M. E., F. Lario, A. Ortiz, and F. Gómez. 2013.“Available-to-Promise Modeling for Multi-Plant Manufac-turing Characterized by Lack of Homogeneity in the Prod-uct: An Illustration of a Ceramic Case.” AppliedMathematical Modelling 37 (5): 3380–3398.
Arshinder, A. K., A. Kanda, and S. G. Deshmukh. 2008. “Sup-ply Chain Coordination: Perspectives, Empirical Studiesand Research Directions.” International Journal of Produc-tion Economics 115 (2): 316–335.
Azevedo, A. L., and J. P. Sousa. 2000. “A Component-BasedApproach to Support Order Planning in a DistributedManufacturing Enterprise.” Journal of Materials ProcessingTechnology 107 (1–3): 431–438.
Balakrishnan, A., and J. Geunes. 2000. “Requirements Planningwith Substitutions: Exploiting Bill-of-Materials Flexibilityin Production Planning.” Manufacturing & ServiceOperations Management 2 (2): 166–185.
Bhakoo, V., P. Singh, and A. Soha. 2012. “CollaborativeManagement of Inventory in Australian Hospital SupplyChains: Practices and Issues.” Supply Chain Management:An International Journal 17 (2): 217–230.
Bititci, U., T. Turner, D. Mackay, D. Kearney, J. Parung, andD. Walters. 2007. “Managing Synergy in CollaborativeEnterprises.” Production Planning & Control 18 (6):454–465.
Boza, A., A. Ortiz, and L. Cuenca. 2010. “A Framework forDeveloping a Web-Based Optimization Decision SupportSystem for Intra/Inter-Organizational Decision-MakingProcesses.” Balanced Automation Systems for FutureManufacturing Networks 322: 121–128.
Framiñán, J. M., and R. Leisten. 2009. “Available-to-Promise(ATP) Systems: A Classification and Framework forAnalysis.” International Journal of Production Research 48(11): 3079–3103.
Gomes da Silva, C., J. Figueira, J. Lisboa, and S. Barman.2006. “An Interactive Decision Support System for anAggregate Production Planning Model Based on MultipleCriteria Mixed Integer Linear Programming.” Omega 34(2): 167–177.
Hernandez, J. E., R. Poler, J. Mula, and F. C. Lario. 2011. “TheReverse Logistic Process of an Automobile Supply ChainNetwork Supported by a Collaborative Decision-MakingModel.” Group Decision and Negotiation 20 (1): 79–114.
Holweg, M., and F. K. Pil. 2008. “Theoretical Perspectives onthe Coordination of Supply Chains.” Journal of OperationsManagement 26 (3): 389–406.
Jagdev, H. S., and K. D. Thoben. 2001. “Anatomy of Enter-prise Collaborations.” Production Planning & Control 12(5): 437–451.
Kubat, C., E. Öztemel, and H. Taşkιn. 2007. “Decision SupportSystems in Production Planning and Control.” ProductionPlanning & Control 18 (1): 1–2.
Lambert, D. M., and M. C. Cooper. 2000. “Issues in Supply ChainManagement.” Industrial Marketing Management 29: 65–83.
Lejeune, M. A., and N. Yakova. 2005. “On Characterizing the4 C’s in Supply Chain Management.” Journal of Opera-tions Management 23 (1): 81–100.
Okongwu, U., M. Lauras, L. Dupont, and V. Humez. 2012. “ADecision Support System for Optimising the Order FulfilmentProcess.” Production Planning & Control 23 (8): 581–598.
Pibernik, R. 2006. “Managing Stock-Outs Effectively withOrder Fulfilment Systems.” Journal of ManufacturingTechnology Management 17 (6): 721–736.
Poler, R., J. E. Hernández, J. Mula, and F. C. Lario. 2008.“Collaborative Forecasting in Networked ManufacturingEnterprises.” Journal of Manufacturing TechnologyManagement 19 (4): 514–528.
Romano, P. 2003. “Co-Ordination and Integration Mechanismsto Manage Logistics Processes across Supply Networks.”Journal of Purchasing and Supply Management 9: 119–134.
Zschorn, L. 2006. “An Extended Model of ATP to IncreaseFlexibility of Delivery.” International Journal of ComputerIntegrated Manufacturing 19 (5): 434–442.
Production Planning & Control 661
Dow
nloa
ded
by [
141.
212.
109.
170]
at 0
7:16
20
Nov
embe
r 20
14