A B2C e-commerce intelligent system for re-engineering the e-order fulﬁlment process

(1)

ContentslistsavailableatScienceDirect

Expert Systems With Applications

journalhomepage:www.elsevier.com/locate/eswa

A B2C e-commerce intelligent system for re-engineering the e-order fulﬁlment process

K.H. Leung, K.L. Choy

^∗

, Paul K.Y. Siu, G.T.S. Ho, H.Y. Lam, Carman K.M. Lee

Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

a rt i c l e i nf o

Article history:

Received 30 May 2017 Revised 25 July 2017 Accepted 9 September 2017 Available online 11 September 2017 Keywords:

E-commerce logistics O2O retailing Order fulﬁlment

Business process re-engineering Warehouse postponement applications Expert systems

a b s t ra c t

Intoday’s worldofdigitization, therise ofthe e-commercebusiness around theglobe hasbrought a tremendouschangenotonlyinourpurchasinghabits,butalsototheentireretailandlogisticsindustry.

Giventheirregulare-commerceorderarrivalpatterns,limitedtimefororderprocessingine-fulfilment centres,andtheguaranteeddeliveryschedulesofferedbye-retailers,suchassame-dayornext-dayde- liveryuponplacinganorder,logisticsserviceproviders(LSPs)mustbeextremelyefficient inhandling outsourcede-commercelogisticsorders.Withoutre-engineeringtheorderfulfilmentprocesses,theLSPs arefoundtohavedifficultiesinexecutingtheorderfulfilmentprocessduetothetighthandlingrequire- ments.This,inturn,delaysthesubsequentprocessesinthesupplychain,suchaslast-miledeliveryoperations,consequentlyaffectingcustomersatisfactiontowardsboththeretailerandtheLSP.Inviewofthe needtoimprovetheefficiencyinhandlinge-commerceorders,thisstudyaimsatre-engineeringtheful- filmentprocessofe-commerceordersindistributioncentres.Theconceptofwarehousepostponementis embeddedintoanewcloud-basede-orderfulfilmentpre-processingsystem(CEPS),byincorporatingthe geneticalgorithm(GA)approachfore-commerceordergroupingdecisionsupportandarule-basedin- ferenceengineforgeneratingoperatingguidelinesandsuggestingtheuseofappropriatehandlingequip- ment.Throughacasestudyconductedinalogisticscompany,theCEPSprovidesorderhandlingsolutions forprocessinge-commercelogisticsordersveryefficiently,withasignificantreductioninorderprocess- ingtimeand traveling distance.Inturn, improvedoperatingefficiencyin e-commerceorderhandling allowsLSPs tobetter align strategicallywith onlineretailers, whoprovide customerswith aggressive, guaranteeddeliverydates.

1. Introduction

The e-commerce business is rapidly expanding around the globe. To increase physical presence, online retailers are using brick-and-mortarstoresforthedisplayoftheir products (referred toas“showrooming”)whileenablingcustomerstomakepurchases online. The rise of such online-to-oﬄine (O2O) retailing and e- commercebusinesshasrevampedtheentireorderfulﬁlmentpro- cessalongsupplychains(Lekovic&Milicevic,2013).Arecentstudy by Forrester Research (2016a) indicated that the total online re- tailrevenuesinChina,Japan,SouthKorea,India,andAustraliawill nearly double from US$733 billion in 2015 to US$1.4 trillion in 2020. While the emerging e-shopping trend is expected to con- tinue,bringingbillionsoftransactionstoeverycorneroftheworld,

∗ Corresponding author.

E-mail addresses: [email protected] (K.H. Le- ung), [email protected] , [email protected] (K.L. Choy), [email protected] (P.K.Y. Siu), [email protected] (G.T.S. Ho), [email protected] (H.Y. Lam),[email protected] (C.K.M. Lee).

order fulfilment along supply chains has been identified as one ofthe major bottlenecksaffecting the developmentof theglobal e-business and endconsumers’ online purchaseexperience (Cho, Ozment, & Sink, 2008; Wang, Zhan, Ruan, & Zhang, 2014). On the one hand, logistics serviceproviders (LSPs) attempt to grasp the blooming market segment of e-commerce in a wider business perspective. Onthe other hand,however, atthe operational level,they arestruggling withthe problemofe-commerce order handlinginefficienciesinwarehousesordistributioncentres(Lang

& Bressolles, 2013). Though the bottlenecksof e-order fulﬁlment process, asillustrated in Fig.1, havereceived greater concern by bothindustrypractitionersandresearchersinrecentyears,thein- creased complexity anddynamism of the handling requirements ofe-commerceorders,i.e.ordersbeingpurchasedviatheInternet usingcomputersor smartphones,haveworsened thelogisticsin- dustry’sheadaches. This phenomenon is largely attributedto the difference between e-commerce orders andtraditional orders, as showninTable1(Leungetal.,2016).

Traditionally, logistics orders processed in warehouses are mainly initiated from retailers who require stock replenishment http://dx.doi.org/10.1016/j.eswa.2017.09.026

(2)

Fig 1. Order fulﬁlment under B2C e-commerce business environment.

Table 1

A comparison between the nature of traditional logistics orders and e-commerce orders ( Leung et al., 2016 ).

Order characteristics Traditional logistics orders E-orders placed by end customers electronically

Order arrival Regular Irregular

Order nature Mostly stock replenishment Fragmented, discrete

Size per order In bulk In small lot-size

SKUs involved in each order Very few or even identical Many

Number of orders pending for processing Less, relatively easy to predict More and unlimited, relatively diﬃcult to predict

Time availability for fulﬁlment Less tight Very tight

Delivery schedule Relatively more time buffer Next-day or even same-day delivery

for specified physical stores. In contrast to these traditional orders, generallyinvolving only a few typesof stock-keeping units (SKUs), but in large quantity, the business-to-consumer(B2C) e- commerceorders,whichareplacedbyendconsumers,aresignifi- cantlymorewide-spreadintermsofdeliverylocationandinvolve alargenumberofSKUs,butwitheachSKUdemandingonlyavery small quantity. Adding the requirementof same-day ornext-day deliveryfore-commerceorders,thee-commercelogisticsbusiness model is now more complex and dynamic than we could have imagined. The typical order fulfilment process under the B2C e- commerce business environment is illustrated in Fig. 1. Previous studies have already addressed the difficulty of managing warehouse operations and last-mile order fulfilment, without strate- gic and operational transformation, by logistics service providers (LSPs)(Choetal.,2008;Hultkrantz&Lumsden,2001;Lang&Bres- solles,2013).Therefore,LSPs, who capturethee-commercelogis- tics business, can no longer follow the conventional orderfulfil- mentprocesstohandlee-commerceorders.Withouttheorderful- filment process being re-engineeredfor today’s e-business, there are twosignificant problemsinthe existingoperations, asshown inFig.1;theyare:

(i) Alackofmechanismfordatapre-processingofe-commerceorders

In the case of Hong Kong, being a global transhipment hub, LSPsareshiftingtheir businesstoan e-commerceorientation ow- ing to the fast growing trend of e-business in Asia. Mostof the logisticspractitioners,however,lackaneffectivemechanismfore- commerceorderpre-processing,astheir operationsarestill man- ualandwithoutITsupport.E-commerceordersarehandledinthe conventionalwaytothatoftraditionallogisticsorders.

(ii) Ineﬃciencyofe-commerceorderhandlingduetofrequentanddis- cretearrivaloforders

Thefrequentanddiscretearrivalofe-commerceorders,oneof thebiggestdifferencesascomparedtothetraditionallogisticsor- ders, has resulted in e-commerce order handling in e-fulfilment centres being inefficient. There is a lack of lightweight, cost ef- fective IT solutions that are specifically designed to handle e- commerceorders which are receivedfrom theInternet. The core reason for this is the lack of domain know-how by IT solution developers in the e-commerce supply chain field, thereby being unabletoidentifytheB2C e-commerceorderhandlingdifficulties currentlyfacedbythelogisticspractitioners.

Consequently,theabsenceofamind-setinthemanagerialper- spective andan effectivemechanism in the operational perspec- tivefororderpre-processinghascreatedbarriersforlogisticsprac- titionersto engageinthe e-commerce logisticsbusiness. Further, not only does the internal incapability of efficient e-order han- dlingbecomethemajorobstacleinbusinessexpansion,butisalso abottleneckintheentiree-commercesupplychainwhichaffects theefficiencyofe-fulfilmentofthedownstreamsupplychainpart- ners.Thisexplainswhythelast-miledelivery ine-commerce,the final leg of the complete journey of a parcel before it reaches the customer, is regarded as one of the biggest challenges in today’s e-commerce business. In view of the necessity of logistics process re-engineering under the emerging e-commerce logistics business environment, this paper presents a cloud-based e-orderfulfilment pre-processingsystem(CEPS),which integrates the concept of warehouse postponement, an operational strat- egyof“groupingpendinglogisticsorders forprocessinginbatch”

(Leungetal.,2016).Throughtheincorporationofthegeneticalgo-

(3)

Table 2

A summary of the literature related to warehousing and transportation activities.

Scope Researchers

Warehouse layout design Yao et al., 2010 ; Hassan, 2002 ;

Caron, Marchet, and Perego, 20 0 0 ; Önüt, Tuzkaya, and Do ˘gaç, 2008 Storage location assignment in warehouses Yang et al., 2015 ;

Pan et al., 2015 ; Chew and Tang, 1999 ; Jane, 20 0 0 ;

Muppani and Adil, 2008

Order picking time reduction De Koster, Le-Duc, and Roodbergen, 2007 ; Petersen, 20 0 0 ;

Bindi, Manzini, Pareschi, and Regattieri, 2009 Resource management in warehouses Chow, Choy, Lee, and Lau, 2006

Design of warehouse scheduling system Zacharia and Nearchou, 2016 ; Chan and Kumar, 2009 ; Park, Kang, and Lee, 1996 Transportation routing and scheduling Zegordi, Abadi, and Nia, 2010 ;

Chen and Lee, 2008 ; Geismar et al., 2008 ;

Taniguchi and Shimamoto, 2004

rithm(GA)approachfore-commerceordergroupingdecisionsup- port,arule-basedinferenceengineforgeneratingoperatingguide- lines andsuggesting the use ofappropriate handling equipment, LSPsare ableto consolidatepending,discrete e-commerceorders forbatch releaseandprocessing, thereby improvingtheeffective- ness in e-order handlingand meeting the same-day delivery re- quirementsofe-commerceorders,even inthe caseofreceivinga large numberof orders simultaneously. The contributions of this paperaretwofold.First,thisstudycontributestoawiderbody of theliterature,whichhasrarelyaddressedtheoperatingissuesand bottlenecksspeciﬁcallyinhandlinge-commerceorders.Amajority oftheexpertsystemsproposed inprevious studiesinthedomain ofwarehousing andtransportation process improvement,such as Accorsi,Manzini,andMaranesi(2014),Guetal.(2016),Lam,Choy, andChung (2011),Oliveira,Cardoso,Barbosa, daCosta,andPrado (2015), Patriarca, Costantino, and Di Gravio (2016), Poon et al.

(2011),Yang,Miao,Xue,andYe(2015),Yao,Paik,andWedel(2010), Zachariaand Nearchou (2010), Taniguchi andShimamoto (2004), ChanandKumar(2009),andChenandLee(2008),focusontack- lingaspecificoperationalissueinwarehousesordistributioncen- tresinhandlinggenerallogisticsorders.However,withoutconsid- erationofthedifferencesinthenatureandhandlingrequirements betweene-commerceordersandconventionallogisticsorders,previous expert systems might not be applicable to the scenario of today’se-commerceorder handlingprocess.Therefore, thispaper fillsthisgapintheliteraturebyproposingadecisionsupportsys- temforstreamlininge-orderfulfilment,therebyprovidinginsights forfutureresearchinlogisticsbusinessprocessre-engineeringun- dertheeraofe-commerce.Second,theproposed hybridsolution, whichintegrates GAand rule-basedinference engine, groupsthe orders forpicking at the sametime to minimize repeat visitsto nearbystoragelocations. Thisstreamlinestheorder pickingoper- ationsof itemsintheir storagelocationsinwarehouses ordistri- butioncentres,one ofthecostliest activities amongst thevarious logisticsoperatingcategories,i.e.receiving,storage,pick-and-pack, anddelivery.

Therestofthispaperisorganizedasfollows.Section2presents aliteraturereviewoftheexisting researchandtheoriesrelatedto the topic. The CEPS model design is presented in Section 3,followedby acasestudytodemonstratetheimplementationproce- duresofthemodelinSection 4.Theresultsandﬁndingsare dis- cussedin Section 5. Finally, Section 6 containsthe conclusion of thestudyanddirectionsoffutureresearch.

2. Literaturereview

In a traditionalsupply chain, goods are processedin a multi- level supply chain inorder to transport thegoods fromthe fac- torytophysicalretailstores. Endconsumers makepurchasesand receive the products at physical stores. In today’s Omni-channel retailing, the buying process of the end consumer involves various sources, from online to offline. End consumers’ orders can be received anytime and anywhere by the e-retailer. As orders are placed via the Internet, the downstream of the e-commerce B2C supply chain consists of a large number of unknown desti- nations spread around the world, requiring direct home delivery orconsumer-direct delivery.The underlying fulfilmentoperations for e-commerce shipments, also callede-fulfilment (Agatz, Fleis- chmann, & Van Nunen, 2008), is a crucial driver of e-commerce growth(Maltz, Rabinovich,& Sinha,2004; Morganti, Seidel,Blan- quart,Dablanc,&Lenz,2014).Whilelogisticsanddistributionplay essentialrolesinthee-commercesector(Chen&Lin,2013;Esper, Jensen,Turnipseed,&Burton,2003),logisticspractitionersengaged ine-business havebeenfacingavariety ofchallengesinfulfilling onlineB2Ccustomer ordersduetothe e-fulfilmentprocessbeing fundamentally different from the traditional shipments handling process in termsof the ordernature and handlingrequirements, inventorymanagement, warehousedesign andmanagement,last- mile delivery, and returns management (Agatz et al., 2008; De Koster,2003; Fernie& McKinnon,2009; Leungetal., 2016;Maltz etal.,2004).

Theorderfulfilmentprocessintraditionalwarehousesincludes four major aspects: order receiving, order storage, order picking andpacking,andorderdelivery.Amongstthefourcategoriesofor- derhandlingoperations,order-pickingisthemostlabourintensive operation in the warehouse and induces the highest warehouse- associated costs (Accorsiet al., 2014). Ine-fulfilment, the operating categoriesin warehousingand distributionare commonwith respecttotraditionalorderfulfilment.However, order-pickingop- erations in e-fulfilment centres are initiated by end consumers whoplacedordersrequiringthee-retailers orthelogisticsservice providers to fulfil the orders accordingly. Such a demand-driven distributionmodelintheeraofe-commercefurtherincreasesthe complexity of order picking operations, asthe order arrival pat- ternin moredifficult topredict, comparingto conventional large lot-sizedlogisticsorders forweekly orbi-weeklystockreplenish- ment ofdesignatedretail stores. Therefore,the importanceof lo- gisticscapabilityandoutsourcingislikelytoincrease,requiringan

(4)

Fig. 2. E-order fulﬁlment process with CEPS.

entirelynewfulﬁlmentinfrastructuretohandlee-commerceship- ments (Chan,He,&Wang,2012;Cho etal.,2008; Morgantietal., 2014;Xing,Grant,McKinnon,&Fernie,2011).

Arangeofresearchactivitiesregardingwarehousingandtrans- portationactivitiescanbefoundinthemainstreamliterature,and a summary isshown inTable 2,forthe areasspecifically related tothedesignandprocessimprovementinwarehouses.Concerning the warehouses anddistributionoperations, streamliningthe traditional orderfulfilment process through providing decision support for logistics practitioners has become one of the active re- searchareas.Poonetal.(2011)integratedradiofrequencyidentifi- cation(RFID)technologywiththegeneticalgorithm(GA)technique forgeneratingpick-upanddeliveryrouteplansforsmallbatchre- plenishmentorders.Lametal.(2011)proposedadecisionsupport system integrating the case-based reasoning (CBR) techniquefor supporting managersin making appropriate order fulfillingdeci- sions.Thoughvariousaspects ofthewarehousingandtransporta- tion sector have beenwidely examined, previous research activi- tiesfocused ontraditional warehousingoperationsor transporta- tionoperations,andtheattentionpaidbyresearchersinconsider- ation of today’se-commerce orderfulfilment inthe warehousing andtransportationsectorisverylimited.

The usedofknowledge-basedtechniquesinsolving arangeof decision-making problems in manufacturing and logistics activi- tiesis commonnowadays(Hoetal., 2008; Tana,Limb, Plattsc, &

Koay, 2006). Thegenetic algorithmtechnique, one ofthepopular artiﬁcial intelligence (AI) techniques in tackling real-world prob- lemwithoutrequiringhugecomputationaleffort,hasbeenproven toexcelinsolvingcombinatorialoptimizationproblems(Hoetal., 2008).The GAoperation isbasedontheprinciplesofgeneticand naturalevolution throughrandom selection andthereproduction ofoffspring(Lee,Choy,Ho,&Lam,2016;Renner&Ekárt,2003).A nearlyoptimalsolutiongeneratedbytheGAtechniqueisinaform ofchromosome,whichinvolvesastringofgenes.Twobasicoper- atorsinGA, thecrossoveroperator andmutationoperator, create new offspring from the parent chromosomes by selecting a pair of chromosomes for matching,and performing a random adjust- ment ofsomevalues ofthegenes inachromosome,respectively.

The generated chromosomes are then beingevaluated through a deﬁned ﬁtness function.A number ofGA applications havebeen foundinthedomainofmanufacturing,warehousinganddistribu- tion, showing theuse of GA forsolving scheduling-related problems.Lin,Choy,Ho,andNg(2014)developedageneticalgorithm- basedoptimizationmodelformanaginggreentransportationoper- ations. Mendes,Gonalves,andResende(2009)integratedGAwith heuristics rules for tackling the joint replenishment problem in warehouses.Leeetal.(2016)providedacomprehensivequalityas-

surancescheme in the garment industry through optimizing the fuzzyrulesusingageneticalgorithm.

Numerous intelligent systems and approaches for providing order-handlingdecision support in warehouses have been devel- oped. However, there is a scarcity in the literature on B2C e- commerce order fulﬁlment in distribution centres, which takes thee-orderhandlingprocessandrequirementsintoconsideration.

Onlya limitedrangeofpapersrelatedtoB2Ce-commercecanbe foundintheliterature,such asassessingthelevelofB2C trustin e-commerce(Akhter,Hobbs,&Maamar,2005),developinganego- tiationmodelforB2C ecommerce(Huang,Liang,Lai,& Lin,2010), aprototypeofe-commerceportal withasetofservices provided byintelligentagents(Castro-Schez,Miguel,Vallejo,&López-López, 2011), and an evaluation model for ranking B2C websites in e- alliance (Yu, Guo, Guo, & Huang, 2011). Hence, in this paper,an e-orderfulfilmentpre-processingsystemisproposed,whichhigh- lightstheimportance of usingthe geneticalgorithm approachas the core means of tackling the common operational bottlenecks foundin e-fulfilmentcentres, withan integrationofa rule-based inferenceengineforfurtherprovidingamorecomprehensivesolu- tiontoassiste-commerceorderfulfilmentoperations.

3. Cloud-basedE-orderfulﬁlmentpre-processingsystem

Inordertomeetthetightdeliveryrequirementsofe-commerce orders,theoutsourcede-orderfulfilmentprocessmustbeexecuted efficiently.Thisrequires warehouseprocess re-engineeringby the logistics service providers. Therefore, a cloud-based e-fulfilment planning system (CEPS) is proposed, for streamlining and re- engineeringtheflowofe-orderfulfilmentoperations.Throughthe useofCEPS,thedecisionsupportprovidedbytheproposedsystem enableslogisticsoperatorstopostponeorderfulfilmentprocessex- ecutionintheire-fulfilmentcentresbyeffectiveconsolidationand planningofincominge-ordersforlaterbatchprocessing,insteadof conventionally processingeach e-order individually, asillustrated inFig.2.CEPSisawebapplication(orwebapp)whichcomprises twomodules,namely(i)E-orderinformationcollectionandsorting module,and(ii)E-ordergroupingandresourceallocationmodule.

Fig.3depictsthearchitectureofCEPS.

3.1. E-orderinformationcollectionandsortingmodule

Inthe eraofthe e-commercebusiness, delivery ordersplaced by end consumers are retrieved from the Internet. Eﬃcient retrieval and consolidation of an order therefore requires a cloud databaseintegratedintoawebappforrealtimedataretrievaland processing.The front-endofthismoduleinvolvesauserinterface (UI)intheCEPS’sweb app,allowingusers,typicallythecustomer

(5)

Fig. 3. Architecture of cloud-based e-order fulﬁlment pre-processing system speciﬁcally designed for e-commerce order handling.

service (CS) staff in a logistics company, to retrieve the details ofe-orders pending forfurther processing andconfirmation, and tomanually update them ifnecessary. The web app, which con- sistsofa seriesofweb pages, isconstructed by HypertextMark- up Language(HTML). Any action madeby the userson the web pageswouldtriggeranupdateontheclouddatabaseofCEPS.The databaseofCEPSistheinformationrepositoryforcollecting,stor- ingandsortingtwotypesofdata:(i)deliveryorderdetails,which are received in real time via the Internet either from e-retailers ordirectly from endconsumers, and (ii) the basicsettings ofe- fulfilmentcentre,whicharestaticinformationpreliminarilystored inthecloud databaseforretrieval.The detailsofthesetwomajor typesofdatastoredintheclouddatabasearedisplayedinTable3. The major data processingoperations inthis moduleincludes databasequeryprocessing, datasorting anddisplay.Fordatabase query processing, essential data as shown in Table 3 for insert, view,edit, delete andupdate can be performedin the UI of the CEPSthrough a setof structure querylanguage (SQL) statements designedandstoredintheSQLdatabase.Fordatasortinganddis- play,the operation is done automatically in the back-end ofthe databaseso that all retrievede-orders are aggregatedand sorted bystock-keepingunits(SKUs),disregardingwhichparticularSKUs are fulfilling which customer order. An illustration is shown in Fig.4.Withtherearrangedorderinformation,alistofitemstobe

processedine-fulﬁlmentcentreisdisplayedintheUIoftheCEPS, which serves as the input of the subsequentmodule for e-order groupingandresourceallocationdecisionsupport.

3.2. E-ordergroupingandresourceallocationmodule

The sortedorderdetails andstoragelocation ofSKUs pending to be picked are the inputs of this module. The grouping of e- ordersusingtheGAmechanismstartswithencodingtheproposed ordergroupingmodelintoachromosome.Aninitialpopulationof e-ordersgroupingsolutionisthenformed,followedbythefitness ofeachchromosomebeingevaluatedwiththeadoptionofaquan- titative model that includes constraints as the representation of theorder groupingcriteria. Priortoreaching the terminationcri- teria,crossoverandmutationoperationsarerepeatedlyperformed to generate different sets of solutions. Upon fulfilling the termi- nationcriteria,the chromosomewithsmallestfitness valueisse- lected as the near-optimal solution, which will then be decoded andtransformed into a complete e-order grouping plan withre- sourceandoperatingguidelinesgeneratedforeachordergrouping list.Anoverviewoftheprocedures inmodule2ofCEPSisshown inFig.5.

(6)

Table 3

Generic details of the information stored in CEPS’s cloud database.

Types of data:

(i) Customer order details

Details Data type Data source

Ordered item (presented as SKU no.) Numeric Real time retrieval from retailers of end consumers

Item quantity Numeric

Item weight Numeric

Order time Date

Estimated time of delivery (ETD) Date

Delivery location String

Order number Numeric

Order priority String

Customer ID Numeric

(ii) Initial setting of e-fulﬁlment centres

Storage location setting (Zone and bin level) String Initial input as part of the construction of the cloud database Travel distance between each bin location Numeric

Storage location of each SKU String

Equipment master String

Fig. 4. Consolidation and sorting of customer e-order in CEPS.

3.2.1. Chromosomeencoding

The chromosome inthe CEPS isa solution foridentifying the nearlyoptimalcombinationsofSKUstobepickedunderthesame ordergroupinglist.Severalordergroupinglistswillbeformedand presentedinthechromosome.AsshowninFig.6,thegenericfor- matofachromosomeisdividedintotwoareas:ordergroupingre- gion,andparameterregion.Fortheordergroupingregion,theba- sicideaofthechromosomeencodingschemeofthisregioncomes from Lin etal. (2014). The value of each gene is a real number, where“0” representsthedepot,andothervaluesindicatethestor- age bin location. Take the chromosomein Fig. 6 as an example, a chromosome “112 114 145 0 243 231 212 321 0 410 412 413 4150” canbeinterpretedas0-112-114-145-0,0-243-231-212-321- 0,and0-410-412-413-415-0,whichimpliesthatatotalofthreeor- der grouping listsare generated, withthe storagelocation under each order groupinglistspeciﬁed inthe chromosome.Forexam- ple, the order grouping list with chromosome “0-112-114-145-0”

denotes thatthreestoragebinlocations,i.e.112,114and145,are tobetravelledtowhenoperatorinthee-fulﬁlmentcentrefollows thisordergroupinglistinexecutinge-orderfulﬁlmentoperations.

The parameter region can further be classiﬁed into three different areas:totalweight, totalquantity andrequiredequipment.

Thegeneticoperationswillbeperformedonlyintheordergroup-

ingregion,astheparameterregiondisplaysthecorrespondingto- talweight (W), totalquantity (Q) andtypesof SKUsinvolved (S) ineachordergroupinglistbasedontheordergroupingsolutions generatedinthechromosomegenesintheordergroupingregion.

Theinformationprovidedintheparameterregionservesasthein- putoftherule-basedinferencesystem,whichgeneratesoperating guidelinesandsuggeststherequiredequipmentbasedonthepre- deﬁnedrules.AsshowninFig. 6,thecorresponding totalweight, totalquantityandrequiredequipmentfortheordergroupinglist1, i.e.chromosomegeneB_1j,wherej=1,2,…,N,aredenotedasW₁, Q₁,S₁ respectively.Thelengthofthechromosomeintheparame- terregiondependsonthenumberoforderpickinglistsgenerated forthecurrentproblem.

3.2.2. Populationinitialization,ﬁtnessevaluation,andgenetic operations

Duetoa largenumberofrequiredstoragelocationstobevis- itedinane-fulﬁlmentcentrefororderpickingofvariousitemsor- deredbydifferentonlinecustomers,thelonglengthofachromo- some suggeststhat a large population size is requiredforgener- atingaconsiderablenumberofpossiblecombinationsincrossover andmutationoperationsintheGAmechanism.Aﬁtnessfunction that minimizesthe one-dimensional travel distancebetweentwo

(7)

Fig. 5. E-order grouping and operating guidelines generation in CEPS.

Fig. 6. The generic format of a chromosome.

adjacentnodes,i.e.storagebinlocation,isusedtoevaluatetheﬁt- ness ofeach chromosome.Considering the placement andalign- ment ofa series ofpallet racksin parallel as a common facility layoutinthestoragearea ofe-fulﬁlmentcentres,warehousesand distributioncentres,adistancematrixthatcalculatesandindicates theinter-bindistancesamongeachstoragebinisproposed,instead ofa conventionalcomputationofthedistancebetweentwoadja- centnodesi andi+1 withcoordinates (x_i,y_i) and (x_j, y_j) using D_i,j=

(^xj−x_i)²+ (^yj−y_i)²^.^Thepreparationofaninter-bindis- tancematrixisconsidered tobe a morerationalandappropriate approach,incomparisontoasimplecalculationofthedistancebe- tweentwo adjacentnodesusingtheir x,ycoordinates,duetothe fact an operator who is about to visit two storagebin locations, saybinA1andbinB1inFig.7,mustbeeitherpickingroute1or route2asillustratedinFig.7.Adistancematrixthatillustratesthe shortestpossibletraveldistanceamongeachbinthereforeachieves abetteraccuracyforﬁtnessfunctionevaluation.Theshortestinter- bintraveldistance(D_i,j)forbiniandjiscalculatedbyEq.1: D_i,j=min

X_i+X_j,

(

^L⁻^Xi

)

⁺

L−X_j

+Ns×S+NA×A (1)

whereX_iisdistancebetweenthex-coordinateofstoragebiniand thestartingposition oftheaisle(the origin),Listhetotallength ofan aisle, N_A andNs are respectivelythe numberof aisles and

Fig. 7. An example of the shortest inter-bin travel distance calculation between two storage bins.

storagebinsverticallytravelledacross,andAandSarerespectively thewidthsofanaisleandastoragebin.

Foreach order groupingproblem, n out ofN storagebins are requiredtobe visitedtopickuptheitemsorderedbyonlinecus- tomers, where Nis the total number of storagebins, and n≤N.

WiththeN×Ndistancematrixthatcontainsallinter-binshortest travel distances inthe e-fulﬁlmentcentre,the inter-bin distances amongn out ofNstoragebins arerequiredto beextractedfrom

(8)

Table 4

Notation table for quantitative model of CEPS.

Notation Deﬁnition

G Index set of all order grouping lists, G = {1,2,…,g}

N Index set of all storage bin location, N = {1,2,…,n}

g Index for order grouping list i, j Index for storage bin location

v _i Volume of items to be picked at storage bin location i w i Weight of items to be picked at storage bin location i V g Volume limit of order grouping list g

W g Weight limit of order grouping list g

D _ij Shortest inter-bin distance between storage bin location i and j x ijg Binary variable indicating whether order grouping list g ∈ G

travels storage bin location i and j

theparentdistancematrix,by filteringouttheinter-bindistances ofall storagebinsthatwill notbe visitedinthisorderfulfilment wave.ThisallowstheGAmechanismtoevaluatethefitnessofeach chromosomeby onlycopingwiththeinter-bindistances ofthen storagebinsconcernedinthecurrentproblemusingthen×ndis- tancematrixextractedfromtheoriginalN×Ndistancematrix.The quantitative formatofCEPS fore-order groupingispresentedbe- low.ThenotationsaredepictedinTable4.Eq.(2)determinesthe shortesttravel distance ofall generated chromosomes.Constraint (3)ensures thatthe ordergroupinglistincludes thevisitingstor- age bin location j immediately after storage bin location i. Con- straints (4)and(5)ensurethat eachtravel pathonly hasoneor- dergroupinglistandeachstoragebinlocationisincludedinonly onesingleordergroupinglist.Constraints(6)and(7)respectively specifythevolumeandweightlimitofanordergroupinglist.Con- straint(8)ensuresthecontinuityofpath.

Minimize N

i=1

N j=1

G

g=1

Di jxi jg (2)

x_{i jg}=

₁_, _if_order_grouping_list_g_includes

visitingbinlocationjjustafteri

0, otherwise

∀

^g^∈ ^G ⁽³⁾

i∈N

g∈G

x_{i jg}=1,

∀

^j ∈ N (4)

j∈N

g∈G

x_{i jg}=1,

∀

ⁱ^∈ ^N ⁽⁵⁾

i∈N

j∈N

v

ixi jg≤Vg,

∀

^g∈ G (6)

i∈N

j∈N

wixi jg≤Wg,

∀

^g∈ G (7)

j∈N j=i

xi jg=

j∈N j=i

xjig,

∀

ⁱ^∈ ^N,

∀

^g^∈^G ⁽⁸⁾

Through the typical genetic operations in GA, including crossover,mutation,andrepairingoperations,theGAiterationpro- cessisstoppedoncethemaximumnumberofiterationshasbeen reached.Thebestchromosomeisthenselectedasthenear-optimal solutionanddecodedintoareadableformatofordergroupingso- lutions.Thisallowsuserstoobtainthesuggestednumberoforder groupinglistsrequired,andthedetailsofeachordergroupinglist, includingthestoragebinlocationssequencetobe visitedineach ordergroupinglist,andthecorrespondingitemswithquantityin- formationtobehandledateachstoragebinlocation.

3.2.3. Rule-basedoperatingguidelinesdecisionsupport

Rule-based operating guidelinesdecision support is generated inadditiontotheordergroupingsolutions.Astheoperatingpro- cedures andequipment selections vary dependingon the nature of an order, the total weight and quantity of each order groupinglist,asindicated intheparameterregion ofthechromosome, alongwiththeproductcategorieshandledineachordergrouping list,serveasthe antecedents,i.e.the if-partofa rule,ofa setof

“IF-THEN” rulespre-deﬁnedinCEPS,whereasthe requiredequip- mentanda setofsuggestedoperating proceduresare theconse- quent,thethen-partofarule.Among thetwo broadkindsofin- ferenceengines usedinrule-basedsystems,forwardchainingand backwardchainingsystems,theformerone,thatis,adata-driving reasoningstrategy,isadoptedinCEPS,soastoprocesstheknown parameters givenin the parameter region ofthe chromosome to keepusingthe“IF-THEN” rulestosuggestanappropriatesetofop- eratingproceduresandhandlingequipment.

Withtheordergroupingdecisionsupportandknowledgesup- portthroughsuggestinganappropriatesetofoperatingprocedures andequipment, the e-commerce order processingflow inthe e- fulfilment centre is reengineered, as the top management of e- fulfilmentcentrecanflexiblyconsolidatepending alarge number ofdiscrete,smalllot-sizedonlinecustomer ordersandreleasethe jobsinwaveswithclearinstructionsofhowthesejobsaretobe handled.

4. Casestudy

Inordertovalidate theperformance,theCEPSisimplemented in a case company. The case company is a medium-sized Hong Kong-basedlogistics service provider that has specializedin B2C e-commercelogisticsanddistribution servicesinrecentyears.As Chinacontinuestodrivecross-bordergrowth,andits shareofthe onlinecross-bordermarketis expectedtogrow from27%in2015 to 40% in2021 (Forrester Research,2016b), thedeveloping trend ofe-commerce in China has created a golden opportunity in re- centyearsforlogisticspractitionersinHong Kongandwithin the Pearl River Deltaregion to grasp alarge part ofthe e-commerce pieby transformingtheir traditionalB2B logisticsbusinesses into e-commercelogisticsbusinesses.This canbe achievedby provid- ingtotale-logisticssolutionssothate-retailerscanconcentrateon theircorebusinessbyoutsourcingtheentiree-orderfulﬁlment,in- cludinge-commercelast-miledeliveryoperations, tologisticsser- viceproviders.

However, logistics service providers face enormous challenges inthee-commercelogisticsbusinessnotonlyduetothetighthan- dlingrequirementsofe-commerceorders,inwhichitisbecoming popularfore-retailerstoguarantee24-to-48-hourdeliverytocus- tomers,butisalsoaffected bytheinternal ineﬃciency ofe-order handlingandprocessing. The challenges are generic andare also facedbythecasecompany,includingtheheavyworkloadofware- houseoperatorsinfulﬁllingtheordersinatimelymannerandthe increasing frequency of picking and packing wrong items. These typical challenges resultfrom the irregular arrival of e-orders as online customers can place orders at any time via the Internet.

AnotherreasonisduetothefactthateachB2Ccustomerorderin- volvesarelativelylargenumberofvarioustypesofSKUs,although in small quantities, which results in a higher chance of inaccu- rateorderfulfilmentconsidering thelarge numberoffragmented e-ordersthat arerequiredtobefulfilledwithin alimitedtime. In view ofthe operatinginefficiency in managingtheir e-commerce business,the CEPS is implementedin the casecompany withan implementationroadmapasillustratedbelow,whichhighlightsthe essentialstages ofdevelopmentfortheproposed systemtowork onaproductionenvironment.

(9)

Fig. 8. Order picking operations in storage bin locations of e-fulﬁlment centres.

Fig. 9. Computer terminal for e-order consolidation and generating order grouping list.

4.1.Implementationofthecloud-basede-fulﬁlmentplanningsystem

The implementation procedures can be divided into four phases:(i) Clouddatabasedevelopment, (ii) CustomizationofGA mechanism and rule-based inference engine, and (iii) Front-end userinterfacedevelopment.

4.1.1. Clouddatabasedevelopment

E-order consolidation and sorting in the cloud database is one of the most crucial functions of the CEPS for the purpose of logistics process re-engineering in the e-commerce business.

Table5summarizestheessentialfunctionsofthecentralizedcloud databasefordatastorageacrossvariouse-logisticsactivitiesinthe casecompany. The cloud databasecollects e-orders and displays thecollectede-orderspendingprocessing.Thelogisticsorderpro- cessing flow is re-engineered by allowing users to control when pendingorders are readyfor batch releaseto thewarehouse department so as to perform the order fulfilment operations. The centralized relational database also sorts the received orders by SKUs,so that theSKUs tobe handled inthe e-fulfilmentcentres aregroupedforease ofactual processing. Figs.8 and9show re-

spectivelythe storagearea ofthee-fulﬁlmentcentres ofthecase companywhereorderpickingoperationstakeplace,andthecom- puterterminalforthedisplayandconsolidationofe-orders.

4.1.2. Customizationofthegeneticalgorithmmechanismand rule-basedinferenceengine

(i) GAmechanismandthedistancematrix

Togovernthee-fulfilmentoperationalflow,decisionsupportfor groupingSKUs withsimilar storagelocations ine-fulfilment centres, andsuggesting an appropriate set of handling remarks and handlingequipment,isdevelopedthroughtheconstructionofthe proposedGAmechanism. Storagebinlocationsofthee-fulfilment centresaredecodedintorealnumbersforformulatingavalidchro- mosome encoding scheme.AMicrosoft Excelspreadsheetis used forthe back-end algorithm development, andEvolver, a software developedbythePalisadeCorporation,isadoptedtominimizethe fitnessfunctionusingEq.(2)inordertosearchforthebestglobal solution oforder grouping. Fig.10 showsthe ordergrouping de- cisionsupportdevelopmentusinganMSExcelspreadsheet.Adis- tancematrixthatcalculatesalltheinter-bindistancesamongeach storagebinisproposed.Adistancematrixispreparedforthecase companybasedonthelayoutdesignofthestoragebinlocationsin thee-fulfilment centre,asshowninFig.8.Foreach ordergrouping problem, the pending e-orders only involve part of the total numberof SKUsstoredin the storageareas.Therefore, usingthe parentdistancematrix,thatincludesallinter-bindistances,isun- necessary. Inthisregard, a sortingalgorithm,which isdeveloped usingVisualBasicforApplications(VBA),aprogramminglanguage thatautomatetasksinMSExcel,isbuiltforextractingtherequired inter-bindistancesfromtheparentdistancematrixtoformachild distancematrix.The programmingcodeforthe sortingalgorithm isdepictedinFig.11.

(ii) Rule-basedinferenceengine

An appropriate setofoperating guidelinesandorderhandling equipment are suggested through a rule-based inference engine for each of the e-order grouping solutions generated from the GAmechanism. Arule-basedinferenceengineisadoptedforease ofstoringandmanipulating humanknowledge to interpretorder grouping information in a usefulway. The engine is customized for the case company based on the current throughput volume and resource availability for e-order fulfilment operations in the e-fulfilmentcentre.Three parametersare identified asthe factors that influencetheoperatingprocedures andequipmentselection:

total quantity, total volume of the order grouping list, and the typesof SKUsinvolved intheorder groupinglist.An exampleof the “IF-THEN” rules are presented in a decision table format in Table6.IntheCEPS,thewarehousemanagerhastheaccessright to preview all active rules, and add, change, and delete a rule whenever appropriate or necessary. Any change in the database whichstoresthe“IF-THEN” rulesissubjecttoaninternalchecking bythesystemitselfforensuringthereisnoviolationorcontradic- tionamongtherules.

4.1.3. Front-enduserinterfacedevelopment

Afront-enduserinterface (UI)servingasthepresentationand interactiontierforusersisdesigned, asshowninFigs. 12and13. Itintegratestheclouddatabase,back-endGAmechanismandrule- basedinferenceengine,sothatusersareabletoconvenientlyview thenewlyreceivede-orderswhichareretrievedfromtheIntranet andstoredintheclouddatabase;obtaindecisionsupportfromthe GA mechanismregardinghow therequired SKUsfrom thepend- inge-ordersaretobegroupedforbatchprocessingine-fulﬁlment centres;andreceivesuggestionsfromtherule-basedinferenceen- gineregardingtheequipmentandhandlingproceduresofeachor- der grouping listgenerated inthe GA mechanism. The CEPS is a

(10)

Table 5

Database construction for various e-logistics activities.

Related department Logistics activities Functions of the cloud database

Customer service department Customer order inquiry Create log for track and trace of order inquiry history E-order collection Retrieve new orders and stores information in the database Documents for order processing Prepare required documents and import and export, update the

documentation completion status of each order

Shipment notiﬁcation to customers Generate shipment notiﬁcation templates for notifying customers of their orders ready for delivery

Warehouse department Order fulﬁlment in e-fulﬁlment centres Follow the tables for order receiving, put-away, pick-and-pack, and delivery operations

Order status update Update the status of e-order for order visibility and transparency Inventory update Cross-checking and updating of inventory in database and in storage

areas of e-fulﬁlment centres

Fig. 10. Order grouping decision support development using GA.

Table 6

Example rules applied in the case company.

“IF” condition “THEN” action

Example rules for operating guidelines

Quantity is more than 50 Double check if the quantity picked for each SKU is correct at the end of operation

Volume is more than 20 kg Reserve lower space for heavier item

Types of SKUs is more than 5 Pick up item separation tool for separating different SKUs;

Example rules for equipment selection

Volume is 26 –50 kg Use Multi-Storey Trolley for separation of different SKUs Volume is more than 50 kg Use Lifter

web-basedapplicationsothatuserscanloginontothesystemvia the Internet. The Customer Service department can decide when tostop theconsolidation ofe-ordersandstartinitiating theCEPS soastogenerateordergroupingsuggestions,asshowninFig.12. The suggestedoutputs canthen be exportedto other formatsfor modiﬁcation orprintedforthe warehousedepartment toexecute accordingly. The warehouse department can, on the other hand,

modifyoraddvariousinformationtotheCEPS,includingtheavail- ablematerialhandlingequipment,operating guidelinesfordiffer- enttypesofe-ordersandthestoragelocationofSKUs,sothatthe databaseisup-to-date,andthedecisionsupportprovidedbyCEPS isfeasible.

5. Resultsanddiscussion

WithCEPS,thee-fulfilmentprocess inwarehousesordistribu- tioncentresisre-engineered.Logisticsserviceprovidersnolonger performe-orderfulfilmentoperationsimmediatelyafterordersare received online. Instead, e-orders, which are placed by B2C cus- tomersfrom various online salesplatforms, and usually small in lot-size, are consolidated in a cloud database for further order grouping. In addition to the order grouping decision support by separatingpending to-be-pickedSKUsintoseveralordergrouping lists based on storage locations dissimilarity, the operating pro- ceduresandappropriate materialhandlingequipment are further suggestedforeaseofe-orderfulfilmentprocessexecution.Theim- provementisnot onlybeneficial tothecasecompany,butalsoits downstreamlogistics serviceprovidersalong thesupply chain.In this section, the GA optimal parameter setting is first reported, followedby an analysisof thekey performance improvementar-

(11)

Fig. 11. Codes for distance matrix sorting.

easfound inthecasecompany.Finally,the implicationstothee- commercelogisticsbusinessarediscussedindepth.

5.1.Parametersettingsofthegeneticalgorithm

The GA parametersettingsare requiredto be definedpriorto implementationinanactual productionenvironment.Specifically, acrossoverrateandamutationrate,rangingfrom0to1,arede- finedbytheusers.Atrial-and-errorapproachisusedtodetermine anappropriatecrossoverandmutationratethatbestsuitstheGA mechanismdeveloped. Inthis casestudy, a crossover rateof 0.7 and0.9,andamutationrateof0.1and0.25,areselectedfortest- ingunderapopulationsizeof2000andwiththenumberofgener- ationssettobe 50,000.Usingdifferentcombinationsofcrossover andmutation rate as specified in Table 7, it is found that, after pair-wisedexecutingfor10 times,aGA parametersettingwitha crossoverrateof 0.9and amutation rateof 0.1gives thelowest fitnessvalue,assummarizedandshowninFig.14.Themorestor- agelocationstobevisited,agreaternumberoftrialsispreferred, soastoobtainabetternearly-optimalsolution.

5.2.Keyoperatingperformanceimprovements

The performance improvement in the order fulﬁlment opera- tionsine-fulﬁlmentcentrescanbefoundintermsoftwomeasur-

Table 7

GA parameter settings.

Parameter Settings

Crossover rate 0.7/0.9 Mutation rate 0.1/0.25 Population size 20 0 0 Termination criteria Stops if:

(1) the number of generations reach 50,0 0 0; or (2) the target cell has an improvement of less than

0.01% in the last 10,0 0 0 trials.

able areas,they are: (i) Reductionoftotal orderprocessing time, and(ii)Reductionoftotaltravelingdistanceofacustomerorder.

(i) Reductionoftotalorderprocessingtime

The total order processing time in handling an e-order in e- fulﬁllingcentresinvolvesthefollowingsequentialoperations:I.Or- derplanning– Consolidatee-orders andprintrelevantdocuments forwarehouseoperatorstoexecuteaccordingly;II:Picking-Travel tostoragelocationsandpicktherequireditems;III.Packing– Al- locatethepickeditemstothedesignatedcustomerorderandpack theitemsinacartonbox;IV. Labelling– Printandstickrequired label(s)onthepackedcartonbox.

BeforetheimplementationofCEPS,e-orderswereprocessedin- dividually without any grouping of orders in advance for future

(12)

Fig. 12. User interfaces of CEPS – order consolidation and grouping.

Table 8

Order processing time before-and-after comparison.

Operation Before After After (per batch) Difference Order planning 0 min 0.45 min 13.5 min –

Picking 6.8 min 1.2 min 37 min −82.4%

Packing 2.2 min 1.7 min 52 min −22.7%

Labeling 0.33 min 0.33 min 10 min 0%

Total 9.33 min 2.68 min 112.5 min −71.3%

batch processing. Therefore, the order processing operation does not consist of any order planning operation that consolidate e- orders before processexecution. E-orders are immediatelypicked byassigningaworkertotraveltothespeciﬁedstoragelocationsto pick up therequired items. After thepicked itemsare inspected, the items are loaded onto a carton box atthe packing area, fol- lowedbyplacingtheshippingandprecautionlabelsonthepacked cartonbox.Accordingtoaprevious timemeasurement conducted bythecasecompany,thetotale-orderprocessingtimeofanorder, whichinvolvestheoperationsasmentionedabove,is9.33minon average,asshowninTable8.

WiththeimplementationofCEPS,thee-ordersareconsolidated before processing in a batch. On average, each batch consist of 30individualonlinecustomerorders.Through timemeasurement overa 3-monthperiod,fora batch orderprocessingoperation, it takes13.5minfororderplanningoperations, followedby37 min forpicking thegroupede-orders by visitingthe storagelocations once, 52 min for allocating the required itemsto the customers andpacking the30cartonboxes,and10min forlabellingall the packed boxes. On a per-order basis, the total e-order processing timeis2.68minonaverage,showinga70%reductionofthetotal processingtimeascomparedtotheperformancebeforetheimple- mentationoftheproposed system.Agraphicalcomparisonofthe timespentoneachorderprocessingoperationisshowninFig.15. (ii) Reductionoftotaltravelingdistanceofacustomerorder

Thetotaltravellingdistanceisanothernoticeableimprovement area. The adoptionof CEPS hasreducedthe travelingdistance of an e-order as storagelocations are visited only once fora batch of consolidated customer orders, instead of conventionally visit- ingstoragelocationsrepeatedlythroughouttheworkinghoursfor eachparticularcustomerorder.Beforethere-engineeredorderful- ﬁlment operations, the itemspurchased in an e-commerce order

(13)

Fig. 13. User interfaces of CEPS – details of order grouping list and generating operating guidelines.

werepickedbyvisitingthestoragelocationsonce,which,onaver- age,requiresatraveldistanceof68m.Withtheimplementationof CEPS,abatch,whichconsistsof30individualonlinecustomer or- dersonaverage,requiresatotaltraveldistanceof397m.Inother words,only a travel distanceof 13.2 m is required foran order, yielding a81% reduction oftotal travelingdistance ofhandlinga customerorder.Inthelongrun,there-engineerede-orderprocess reducestheworkloadofemployees workingin e-fulﬁlmentoper- ations.The savedtimeforrepeatedvisitstostoragelocationsen- ablesmanagerstoﬂexiblyreallocatethehumanresourcestohan- dle other operations such as put-away andcargo loading or un- loadingoperations.

5.3.Managerialimplications

The riseofe-commerce,O2O retailing,anddirect-to-consumer last-miledelivery haspositioned e-fulﬁlment distribution centres at the very heart of what end consumers perceive asgood service.Aggressive, guaranteeddelivery dates are oftenprovided for addressingcustomer demands. Theincreasing convenienceofon- lineshopping hasredeﬁned the waywe shop; the customer de-

mands, on the other hand, are reshaping e-fulﬁlment. In the e- commerce marketplace, customer segmentation is a knownmar- ketopportunityenablingretailerstoreachawidercustomerbase.

However, it is alsoone ofthe biggest challenges ine-commerce.

Thee-orderfulfilmentoflogisticsserviceprovidershastobevery efficientinhandlinge-orders,whicharereceivedfromtheinternet atanytime,andaretobedeliveredtoavastnumberoflocations worldwidebeforetheguaranteeddelivery dates.Intheabsenceof decisionsupportsystemsfacilitatingthe e-orderinternal process- ingoperationsine-fulfilmentcentres,logisticspractitioners,espe- ciallythoseSME-sizedorganizationsthatoftenhandleorderswith- outcomprehensiveITsupport,experienceobstaclesinmaintaining thesamelevelofefficiencyastheyhadinhandlingtraditionalor- dersinwarehousesordistributioncentres.

The evolution of information and communication technology (ICT) serviceswithcloud computingandmobile technologieshas offeredenterprisesnotonlymoresaleschannelsandeffectivefor- mulation of target marketing strategies through big data analyt- ics, but also better internal and external information and com- municationmanagement solutions forintegrationinto dailybusi- ness for operational excellence. However, with the slow pace of

(14)

Fig. 14. Graphical comparison of the results under different combinations of GA parameter setting.

Fig. 15. Improvement in terms of order processing time.

newtechnology adoptionandinnovationin thelogisticsanddis- tributionsector(Evangelista&Sweeney,2006;EuropeanCommis- sion, 2012), logistics practitioners manage e-orders in a conven- tionalflowofoperations, whichaffectstheire-order handlingca- pabilityandprolongs thee-fulfilmentlead time.Theresults from the casestudy indicates that light-weight IT applications, which integrate artificial intelligencetechniques andthe state-of-the-art cloudcomputingtechnologies,enablelogisticspractitionerstoim- provetheirinternalorderprocessingcosteffectively.Softwareand solutionprovidersshouldtakee-fulfilmentrequirementsintocon- sideration when designing and developing competitive ICT solutions, with theintegration of artificialintelligence techniques for

providing decision support and e-commerce logistics process re- engineeringandautomation.

6. Conclusiveremarksandfuturework

Thecapabilityoflogisticsserviceprovidersine-orderfulfilment isone ofthekey factorsaffecting thegrowthof theonlineretail business.Thispaperdevelopsacloud-basede-orderfulfilmentpre- processing system, which integrates the genetic algorithm technique and the rule-based inference engine, so that logistics ser- viceprovidersareabletoeffectivelyplanfortheupcominginternal processingoperationsofreceivedordersbeforeactualprocessexe- cution.Bysodoing,anywarehousepostponementstrategycanbe realized andsupported by the proposed system, as presented in thispaper.Through consolidating pendinge-orders using acloud database, justifying an optimalinternal orderprocessing plan by thegeneticalgorithmapproach,andprovidingessential operating guidance through the rule-based inference engine for order pro- cessingexecution,logisticsserviceprovidersnolongerhavetopro- cessdiscretee-orders immediatelyaftertheyare received.Thee- commerceinternalorder processingflowis thereforestreamlined andre-designed. The improvede-order handlingcapability oflo- gisticsserviceproviderseventually reducestheprocessingtimein e-fulfilmentcentres,therebymeeting theevertighterdelivery re- quirementsofonlinecustomers.Ultimately,theintelligent system presentedinthispaper contributestothe developmentofthee- commercebusinessenvironmentfromtheperspectiveoftheinter- connectedparties.Logistics serviceproviders become morecapa- bleincapturingthelogisticsofthee-commercebusiness;retailers