Business Process Reengineering driven by customer value: A support for undertaking decisions under uncertainty conditions

Business

Process

Reengineering

driven

by

customer

value:

a

support

for

undertaking

decisions

under

uncertainty

conditions

Yuri

Borgianni

*,

Gaetano

Cascini

,

Federico

Rotini

a_{DipartimentodiIngegneriaIndustriale,Universita` degliStudidiFirenze,ViadiS.Marta,3,50139Florence,Italy} b

FacultyofScienceandTechnology,FreeUniversityofBolzano/Bozen,PiazzaUniversita`,5—Universita¨tsplatz,5,39100Bolzano,Italy

c

DipartimentodiMeccanica,PolitecnicodiMilano,ViaLambruschini,4,20156Milan,Italy

1. Introduction

In rapid changing and highly competitive marketplaces,

industries face the challenge of continuously improving their

offerintermsofproductsandservices.Thedemandforinnovation

reboundsontheindustrialenvironmentandaffectsthebusiness

processes,whichrequirecontinuousandcontrolled updates.On

the one hand, companies are asked to enhance the quality of

productsandservices,soastofulfilthegrowingexpectationsof

customersandstakeholders.Ontheotherhand,firmsstrivetocurb

costsand,generallyspeaking,anyotherchannelledresource.In this

sense,aparamountimportanceisattributedtoallthose initiatives

aimedatstrategicallyredesigningindustrialprocesses inorderto

accomplishsignificantlyhigherperformancesandthat areascribable

tothefieldofbusinessprocessreengineering(BPR). Theliterature

witnessesconsiderableadvantagesarisenbyBPR initiativesand

describestextbooksuccessstories.However,plenty ofcontributions

fromdifferentperiods(e.g.[1,2])pointoutahigh percentageof

unsatisfactoryresultsconcerningBPRpractical implementations,

causingthereforediffusedscepticisminthe field.Recentstudies

providegreaterunderstandingaboutthe successfactorsandmajor

effects ofBPR initiatives,thusadvancing guidelinestogenerate

benefitsfortheenterprisestothegreatest extent[3–6].Thereasonsof

unmetexpectationscanberelatedto disparatecauses.Amongthem,

theliteraturerecognizesthe complexityandthenondeterministic

behaviourofbusiness models[7],aswellastheoverwhelmingfocuson

theminimization ofcosts[8],withtheconsequentdisregardof

workforceinterests andcustomerpreferences.

Received5December2013

Receivedinrevisedform27November2014 Accepted6January2015

Availableonline9February2015

Abbreviations:API,ActivePharmaceuticalIngredient;AT,Attractive;BPR,Business ProcessReengineering;CD,CustomerDissatisfaction;CR,CustomerRequirement; CRM, Customer Relationship Management; CS, Customer Satisfaction; DSS, DecisionSupportSystem;IDEF,Integrated Definition;MB,Must-be;OD, One-dimensional;OS,OverallSatisfaction;OV,OverallValue;POV,PhaseOverallValue; PVA,Process Value Analysis;QFD,Quality FunctionDeployment; VAC, Value AssessmentChart;VE, ValueforExcitingrequirements; VN,ValueforNeeded requirements;VoC,VoiceoftheCustomer.

* Correspondingauthorat:FreeUniversityofBolzano/Bozen,FacultyofScience andTechnology,39100Bolzano,Italy.Tel.:+390471017821;

fax:+390471017009.

E-mailaddresses:yuri.borgianni@unibz.it(Y.Borgianni), gaetano.cascini@polimi.it(G.Cascini),federico.rotini@unifi.it(F.Rotini).

Besides, customer-oriented approaches are largely diffused

withinNewProductDevelopmenttasks.Inthis context,Quality

FunctionDeployment(QFD)playsacentralrolewithinBPR,with

the intended purpose of linking together customer needs and

product design [9]. QFD is aimedat maximizing the perceived

satisfaction of potential clients, on the basis of Voice of the

Customer (VoC) techniques, and indicates through quantitative

metrics the most favourable mix of technical performances.

However,fewQFDapplicationshavebeenexperiencedtorelate

productinnovationdirectionstopractical recommendationsfor

carrying out the redesign of the business processes at the

operationallevel.

Insuchacontext,companiesencounterdifficultiestoselectthe

mostbeneficialinnovationstrategytobeundertakenandthemost

suitableBPRtoolforitsimplementation.Particularproblemsare

experiencedbysmallenterpriseswithlimitedinvestments dedicated

toR&Dactivities.SMEswouldcertainlybenefitof reliableand

easy-to-use Decision Support Systems (DSSs), capable to support the

identificationofthemajorprocesscriticalitiesand thedefinitionof

valuableredesignstrategies.Withthepurposeof overcomingsuch

problems,thepaperproposesasystemfor supportingdecisions,

whose methodological framework has been implemented in a

popularcomputingenvironment(MATLAB R2012b).Thesystem

individuatesthemainweaknessesofa businessprocessandhighlights

themosteffectivedirectionsfor processinnovation.Theproposed

procedure is achieved by radically upgrading and extending a

methodologicalroadmap [10],namelyProcessValueAnalysis(PVA),

swivellingontherole playedbybusinessprocessphasesinfulfilling

customerrequire-ments.Moreinparticular,thebasicmethodhas

beenfurther developedtodealwithuncertaintiesmanagementandto

assess thereliabilityoftheoutputs.Theobjectiveoftheenhancements

is tostrengthentheapproachforundertakingdecisionsconcerning

thechoiceofthemostfavourableBPRinitiatives.

Themanuscriptisorganizedasfollows.Thenextsection presentsa

stateoftheartanalysisfocusedonsystemssupporting decisionsabout

reengineeringactivities.Section3betterexplains thestartingpoint

forthedevelopmentoftheproposalandclarifies themethodological

objectives to be pursued. Section 4 presents the designed

methodological framework and its computer implementation.

Section5showstwoapplicationsofthe methodology.Theformerisan

illustrativecasestudytreatinga manufacturingprocessfromthe

pharmaceuticalfield,carriedout byasampleof27MSEngineering

students.Thepurposeofthis exampleis toshowtheimpactof

divergentopinionsbetween processanalystswithadequatescientific

rigour.Thelatterisareal casestudyfromthefootwearindustry,for

which16individuals haveanalysedthewholebusinessprocessofa

smallItalianfirm withinaprojectofnationalinterest.Thesection

further discusses the emerging outcomes of the experiments.

Eventually,Section6drawsthefinalconsiderationsanddepictsthe

plannedfuture activities.

2. Relatedart

ThisSectionaimsatelucidatinghowtheliteraturetreatsawide

varietyoftopicsrelatedtoindustrialinnovationpractices. According

to the authors’ understanding, process- and product-oriented

approachesseparatelyshowcomplementarybenefits withregardsto

severalaspects.Moreover,inthecomplexnetwork ofexistingsupport

systems,adifficulttaskisrepresentedby weighinguptheefforts

dedicatedtoimprovethequalityof productsandtostemthecostsof

industrialprocesses.The followingSectionsdebatetheaboverecalled

aspectsandoutlinean initialsetofrequirementstodefineanidealDSS

forprocess

reengineering, thusposingthemethodologicalobjectivesofthe

paper.

2.1. OverviewofDSSsforprocessreengineering

2.1.1. GeneralfeaturesofDSSstosupportBPR

BPRinitiativesrepresentcomplexmultidisciplinarytasks, dealing

withmultiplesourcesofrisk[11]andawiderangeof aspectsregarding

differentfieldsofexpertise[12].Furthermore, reengineeringissues

aredirectedtowardscomplexbusiness andindustrialprocesses,

whicharecharacterizedbynot deterministicbehavioursandrequire

dynamic time-dependant models.Theuncertaintyregardingthe

modelandtheparameters governingthebusinessprocessaffectsthe

outputsofBPR tasks.Asaconsequence,firmstendtotakeextremely

risky decisionsintuitivelyratherthanthroughasystematicanalysis.

Itfollowsthatconsistentresearcheffortshavebeendedicated tothe

development of DSSs, modelling instruments and simulation

techniques aimed at increasing the effectiveness of industrial

processes[13]byindividuatingmajorinefficiencies [14].

On the other hand, failures of BPR initiatives can be

explained by strategies oriented on redesigning just the

features pertaining the internal processes [15]. With a

particular insight into process rethinking, it has been argued

[8] that numerous BPR applications have been focused

mainly on resources savings. Such practices, with the

purpose of achieving lean processes by imitating past

experiences, have frequently underestimated the relevance of

the value delivered to customers [16], conversely seen as a

determi-nant for the success of BPR initiatives [17].

The documenteddecision supports that have considered the

supplysideforprocessenhancementhavebeenmostlyaimedat

aligning business strategies towards Customer Relationship

Management(CRM)[18–22].Inthissense,ratherthan

customer-oriented BPR approaches, such DSSs represent methodologies

addressedatimprovingaparticularareaofbusinessandindustrial

management. Eventually, recent contributions exploit customer

feedbacksanddemandstoredesignthebusinessprocess. However,

theseproposalsarerestrictedtospecificindustrial domains,suchas

electronicitemsassembling[23]andservicing [24].

2.1.2. AccountingofuncertaintyissueswithinBPRsystems

Likeanydecision-makingactivity,theredesignandplanningof

businessprocessesisassociatedwithuncertaininputsandrisk. With

referencetosuchaproblemLambertetal.[25]takeinto account

relevant risky factors starting from the modelling phase by

representing such additional information through tailored IDEF

(IntegratedDefinition)frameworks.

ManyresearcheffortshavebeencarriedoutaboutDSSsdealing

withtheuncertaintythatcharacterizesabusinessprocess.Min etal.

[26]developedaDSSsuitableforthebankingindustry, assessing

appropriate BPRtasksunder multi-criteriaanalysis and present

constraints.Williamsetal.[27]dealwithriskand uncertainties

associatedwithBPRinitiatives,focusingonorganizationalhurdlesand

providingguidelinesforpursuingincremental orradicalchangeswith

referencetoexpectedbenefitsand availableinvestments.Wangand

Lin [28] introduced genetic algorithms to efficiently schedule

industrialprocessesforamake-to-ordermanufacturingfirm.Their

researchandapplicationis tailoredforresourceallocationdecisionsin

anenvironment characterizedbytime pressurewithregardsto

deliverydates. Byexploitingsimulationtechniques,Mahdavietal.

[29] builta modelmeanttodynamicallycontroltheproduction

activitiesofa flexiblejob-shop,whereasmanufacturingprocessesare

Still with regards to intelligent decision making

within industrial processes affected by uncertainty,

Gregoriades and Sutcliffe [30] developed a

decision-based system capable to evaluate the advantages of

introducing and managing a new candidate business

process, whose characteristics are known. The system

simulates the business process, by taking into account

industrial performance and human factors, and assesses

oppor-tunities and risks on the basis of the generated

scenarios. Besides, the problem of working with

nondeterministic and uncertain models is compounded by

the presence of qualitative parameters. In such a context,

recent contributions introduce measurable parameters to

deal withuncertainty issues within relevant aspects related to

business processes, i.e. customer relationship [20] and

purchasing management [31].

Overall,theheterogeneousaimsof thesecontributionseither

enhancespecificaspectsofindustrialstrategies,oraretailoredto

supportspecificcategoriesoffirms.Inthissense,theymostlylacka

general and versatile approach, capable tofit theexigencies of

differentindustrialdomainsandencounteredproblems.

2.2. QFDasadecisionsupporttoaddressthedevelopmentofproducts

2.2.1. EmploymentofQFDwithinreengineeringtasks

As already mentioned in the Introduction, QFD represents

the main reference for product development initiatives

stimulated by customer value. The employment of

quantitative variables eases the displaying and the

interpretation of the arising outcomes, thus facilitates decision

making. Further on, QFD shows a robust link with Kano

model [32], thanks to its focus on products features in the

perspective of achieving customer satisfaction. The matching

of QFD and Kano heuristics allows considering the

different impacts of the relevant competing factors on the

perceived value [33,34].

QFDbasicprincipleshavebeenexploitedalsotosupportprocess

redesign.In [35], a modified QFD method replaces engineering

featureswiththefactorscharacterizingthemanufacturingprocess, by

directlyrelatingthelatterwithcustomerrequirements.Themain

achievementofthesystemisthedisclosureofpotentialconflicts

betweendisparateaspectsconcerningthebusinessprocess,rather

thanpracticalhintstosupportdecisionsaboutthereengineering

activitiestobeundertaken.

Ultimately, besides claiming to support a wide range of

reengineering activities, the positive influence of QFD on the

improvementofindustrialprocessesisquestionable[36].Its function

within BPR predominantly consists in supporting the strategic

positioninginthemarketbyanalysingandassessingthe product

performances[37]andchoosingthemostappropriate manufacturing

meansforthedesignedartefacts[38].

2.2.2. Themanagementofuncertainties

Furthermore, a considerabledrawback ofQFD is represented

by diffused uncertainty affecting the inputs and the outputs

of the methodological framework.

Fungetal.surveyedQFDmodels[39]toanalysethereasonsof

uncertaintyintroduction,revealinghowamajorroleisplayedby the

relationships between customer requirements and engineering

characteristics.Inaddition,theirresearchsurveystheeffectiveness of

linearprogramming modelswith fuzzy coefficients tocorrectly

estimatetheextentofsuchrelationships.

The employment of fuzzy set theory represents

the most diffused approach in theliterature for managingthe

uncertainties and the dynamics of the inputs in QFD:

Kahraman et al. [40] proposed a critical review of

these applications. Experiences dealing with uncertainty

carried out by means of fuzzy set theory regard also the

Kano model [41,42], as well as its conjoint utilization

with QFD [43].

Gengetal.[44]introduceafuzzymodelforQFD,takinginto

account, beyond product characteristics, those requirements

involvedin servicesdelivery andpertaining themanufacturing

process.JiaandBai[45]proposeafuzzy-QFDmodeltailoredfor

manufacturingprocesses,whosemainfeaturesareevaluatedby four

industrydomainexperts;theapplicationofthetoolfinally depictsthe

effectsofuncertaininputsinamodifiedHouseof Quality[46].The

surveyedcontributionsarehoweveraffectedby thearguedefficacyof

fuzzysetswithinthemanagementof uncertaintiesfordecisions

undertaking [47]. Furthermore, the relentless difficulties in

employing such complex mathematical models hinder a wide

diffusionofsuchDSSswithinalargeamount ofindustrialcontexts.

2.3. ApproachesforchoosingthemostadvantageousBPRactivities

Accordingtotheaboveoverview,thefieldofdecisionsupport for

BPRappearsasanextremelypopulatedsetoftoolsand methodologies.

Inthisperspective,aconsiderablesupportforthe companieswouldbe

representedbysystemscapableofindividu-atingtheaspectsofthe

wholebusinessrequiringthemost beneficialreengineeringactivities.

Insuchacontext,themostfavourabledirectionsforBPR initiatives

areaddressedbyseveraldecisionsupports,thatapply howeverjustto

peculiarfeaturesofthebusinessstrategy,suchas thetechnicalaspects

oftheprocess[48]orsingleunitsofthe enterprise[49].Reijersand

Mansar[50]provideaframeworkof bestpracticesforBPRtasks

accordingtothefocusofredesign efforts.Heetal.[51]havedevelopeda

FuzzyAnalyticalHierarchy Processtosupportthechoiceamong

differentalternativesof possibleBPRinitiatives.Eventually,Choand

Lee[52]developa web-basedtoolforchoosingthemostsuitable

approach for Business Process Management, according to the

evaluationcriteria dictatedbythecharacteristicsofthefirmunder

investigation.

Amethodologicalapproach,namelyProcessValueAnalysis (PVA)

[10],hasbeendevelopedintheperspectiveofindividuating the

business segments needing major reengineering efforts. The

methodologycharacterizesthemainphasesofanindustrial process

by quantitatively determining their contribution to avoid

dissatisfactionandtoprovideunexpectedvalueforcustomers.It takes

intoaccountalsotheresourcesspenttofulfiltheplanned customer

requirements.Thisallowshighlightingthevaluebottle-necksofthe

businessprocess,soastoaddressthereengineering priorities.Anakin

objective is pursued through the methodology proposed by

JammerneggandKischka[53],thatfocusesnever-thelessjustonthe

enhancementofapeculiarsegmentofthe businessprocesses,i.e.

supplychains.

The PVA simulates the interplay between QFD and Kano

model, although the phases constituting the business

process are considered instead of engineering requirements.

The methodology differs also from the already cited

proposal advanced by Jagdev et al. [35], because it

investigates the single constituent activities and phases of a

business process instead of global characteristics.

2.4. Summaryofthesurveyandindividuationofthemainlackswithin

DSSsforBPR

Table 1 provides an assessment of the examined

approaches with respect to a set of desirable

characteristics for an ideal DSS, gathered from the

above survey. The table highlights, besides the primary

scope of the different proposals, the level at which such

characteristicsareachievedaccordingtotheauthors’ opinion.

With reference to Table 1, the scenario of

proposals for supporting BPR includes tools to prioritize

aspects for product redesign, criteria to individuate

currentprocessandpracticalinstrumentstocarryout

transforma-tionsattheoperationallevel.Suchtoolsresultpoorlylinkedwith

each other and the fulfilment of DSSs capable to guide the

enterprisesinthewholereengineeringprocessstillrepresentsa

severechallenge.Theidealresultwouldbeasystemcapableof

individuatingtheaspectsand theareasofthebusinessprocess

needingmajorchangesandtranslatingtheseinputsintopractical

suggestionstoredesigntheinvolvedindustrialactivities.

Accord-ingtothisobjective,theinitialstepofatargetedresearchactivity

would be the achievement of a DSS module in charge of

supervisingthebusiness processand remarking itsmain

weak-nesses,complyingwiththefeaturesillustratedinTable1.

Suchoutcomemaybeaccomplishedbyextendingtheexisting

contributions (with a specific reference to those closer to the

ultimategoal) beyondthelimitationshighlightedinthesurvey.

According totheauthors’vision,three alternativepathscanbe

followed:

 extendingthepurposeofBPRinstrumentsbasedoncustomers’

opinion;

 improving QFD-like tools tailored to investigate industrial

processes;

 developingtheProcessValueAnalysisfurther.

Table2summarizestheprosandconsenvisionedfortheabove

alternatives.Indetail,thefirsthypothesisregardstheextensionof

the BPR tools swivelling on customer feedbacks, leading to a

methodcapabletoworkindifferentindustrialcontexts.Sucha

method would indicate the most favourable redesign actions

accordingalsototheamountofresourcesrequestedtothefirmfor

Table1

StrengthsandweaknessesoftheexistingdecisionsupportsforBusinessProcessReengineering. Kindofcontribution

tosupportdecision making

Mainaim Kindofoutput tosupportthe decisions (quantitativeor qualitative) Consideration ofthecustomer sphere Consideration ofinternal demands Flexibility accordingto different industrial fields Addressed sectionofthe business process Capabilityto accountof uncertainty issues TraditionalBPR approaches [1,2,11,12] Restructuring industrialprocesses bysuppressing low-valuedactivities Diffusedly qualitative

Moderate Verycareful Modest(due tostandard solutions) Whole Limited BPRapproaches swivelledonCRM Aligningstrategies towardsCRM Diffusedly qualitative

High Varying Varying Justpartially Varying [18–22] BPRsystemsexploiting customerfeedbacks [23,24] Reengineeringbusiness processesaccordingto inputsfromthe customersphere

Diffusedly qualitative

High Limited Verylimited Whole Notrelevant

QualitativeBPR approaches considering uncertainties[25–30]

Varyingaccordingto thesingleproposals

Qualitative Moderate Varying Diffusedlylow Diffusedlyjust partially Foreseen QuantitativeBPR approaches considering Reengineeringstrategic segmentsofabusiness process

Quantitative Moderate Foreseen Limited Justpartially Foreseen

uncertainties[20,31]

QFD(+Kano)[33–35] Aligningproduct profilestomaximize thecustomer satisfaction

Quantitative Verycareful Limited Good Diffusedlyjust theproduct redesign Absent FuzzyQFD(+Kano) [40,43–45] Aligningproduct profilestomaximize thecustomer satisfaction

Quantitative Verycareful Diffusedly absent Good,but complexto beapplied (especiallyin smallcontexts) Justthe product redesign(in mostcases) Foreseen

Systemstochoosethe mostvaluableBPR tool[48–52]

SelectingwhichBPR methodologiescan resultthemost advantageous

Mostly qualitative

Moderate Careful Diffusedlygood Mainlyjusta partofthe business process

Diffusedly limited

ProcessValueAnalysis [10]

Identifyingthe prioritiesforBPR accordingtovalue bottlenecks

Quantitative Careful Careful Good Allthephases composingthe wholebusiness process

Absent

Table2

MethodologicalopportunitiestoachievethefirstmoduleofaDSSmeanttosupervisethebusinessprocessandhighlightthevaluebottlenecks:prosandcons. CandidatestrategytoobtainaDSSmoduleto

highlightprocessweaknesses

Mainadvantagesconcerningitsdevelopment Hurdlestoovercomeinthedesign phaseofacorrespondingmethod ExtensionofBPRtoolsbasedoncustomer

feedbacks

Orientationtowardspracticalmeasurestobeundertaken (easeoflinkingwithsubsequentDSSmodules)

Penaltyforfirmswithoutefficient customerservices

EnhancementofQFDapproachestailoredto industrialprocesses

Developedcapabilitiestomanageuncertaintyissues Arguablepossibilitytoblend differentmethodologies Complexityofsystemsbasedon FuzzySets

FurtherdevelopmentofProcessValueAnalysis Properaccountingoftheresourcesspentbytheindustrialprocess Existenceofpreliminaryindicationsaboutreengineeringdirections

Diffusedemploymentofsubjective evaluations

their pursuance. The advantage of this development strategy

wouldstandinaquicklinkwiththesubsequentmoduleoftheDSS,

sincetheexistingBPRmethodsareconsiderablyorientedtothe

practicalmeasuresforthetransformationoftheprocess.Onthe

otherhand,theresultingDSSwouldsufferfromscarceusabilityfor

firmswithoutdevelopedclientservices,sincethefeedbacksofa

restrictedamount of customers would resultin poorly reliable

indications.

The second development strategy aims at employing the

existing QFD-based approaches, with a particular reference to

thosetoolsinvolvingalsomanufacturingandbusinessprocesses

and already capable to manage uncertainties. A consistent

directionforthedevelopmentstrategy ofQFD-orientedsystems

shouldregard a reliablecomputation of costs and resourcesin

chargeoftheenterprise.Sincethesurveyedproposalsquitediffer

in terms of thefulfilment of the expectedcharacteristics for a

futureDSS,anopportunity(tobehoweververified)isrepresented

byblendingasetofQFD-orientedmethods,attemptingtocombine

thebenefitsofsinglecontributions.Disadvantageswouldconcern

thecomplexityaccountedtosystemsexploitingfuzzysetsandthe

lackofanappropriatestrategytomanageuncertainty.

Eventually,thethirdalternativeconcernstheextensionofthe

ProcessValueAnalysis,whoseintendedpurposefitstheexpected

objectives concerning the recalled initial module of a more

articulatedDSS.Moreover,aninsightfulanalysisoftheapproach

revealsthatitsoutputsincludepreliminaryindicationsaboutthe

most proper BPR practices to be implemented in order to

beneficiallyfollowtheemergingreengineeringdirections.

Accord-ing to the above discussion, the main limitation of this

methodologyconsists in the disregard of uncertainty issues in

viewofanotnegligiblequantityofsubjectiveevaluationsabout

theprocess.AnupgradedversionofthePVA,capabletomanage

uncertainties,wouldrequireanintensetestingcampaigninorder

toassessitsactualreliability.

Despite theabovehurdles,theauthorsdecidedtofollow the

developmentstrategybasedontheextensionofPVA.Thisoption

potentiallyallowsobtainingatoolforintelligentdecisionmaking

tobeemployedwithoutnecessarilyresortingtocostlyand

time-consuming customer surveys. The consequent methodological

objectivestobepursued,betterclarifiedinthefollowingSection,

standin theupgrade of thesystemin terms of accounting for

uncertaintyissues, strivingtosafeguardtheintuitivenessofthe

originalPVAframework.

3. Methodologicalbackgroundandresearchobjectives

Inordertoclarifythemeasuresproposedinthepresentpaperto

develop a DSS with the enunciated scopes, a more detailed

overviewoftheProcessValueAnalysisisherebyprovided.Tables3

and 4 summarize its methodological roadmap, illustrating the

stepsforthedeterminationoftheinputsandthecomputationof

theoutputs,respectively.Saidinputsandoutputsrepresentthe

termsthatsymbolizetheperformancesofthebusinessprocess,its

segments and deliverables. All the steps are featured by an

outcome, i.e. the methodological requirements to be attained,

whichbesidescanbedeemedusefultoqualitativelydescribethe

industrialprocessanditsnuances.ThefollowingSubsectionshows

withgreaterdetailtheoriginalstructureofthemethodologyina

stepwisefashion.

3.1. Descriptionofthereferencemethodology

ThefirststageofthePVAprocedureconcernsthegatheringof

theinformationrelatedtothebusinessprocessunder

investiga-tion.CustomizedIDEF0modelsaresuitabletorepresenttheflows

ofinformationandmaterialsalongtheprocessphases,aswellas

theemployedtechnology,machinery,humanskills.

Complemen-tarydataarecollectedtomaptheexpendituresandthetimingof

eachactivity.TheoverallmodelhelpsindividuatingtheCustomer

Requirements (CRs) that are intended to be delivered and the

organizational and/or manufacturing phases that compose the

businessprocess.

Thenextstepofthemethodologyregardstheinvestigationof

thecustomerrequirementsthathavebeenidentified.According

toauthors’ experience, in orderto performreliable analyses, a

comparabledetaillevelhastoconcerntheschematizationofthe

processintophasesand therepresentationoftheproductsand

services through said customer requirements. Each product/

service attribute is at first characterized in terms of the Kano

categoriescontributingtodelivercustomervalue(One-dimensional,

Attractive,Must-be).Additionally,aswithintheapplicationsofthe

HouseofQuality,relevanceindexesareassignedtotheseattributes,

Table3

InputsofthePVAprocedureandstepsinwhichtheyareemployed,soastogeneratepartialresultswhichcharacterizethetreatedindustrialprocess(columnoutcomes). PVAstep Tasktobecarriedout Outcomes Requiredinputs

1 Informationgathering Processmodel,individuationoftheattributesthatcharacterizethe business,sizingofexpendituresrelevanttoeachphase

Listofphases;listofcustomer requirements(CRs)

2 Evaluatingthereasonsof satisfactionanddiscontentment

CharacterizationoftheCRsaccordingtotheirorientationin determiningexpectedorexcitingquality

Kanocategories 3 Estimatingtheroleplayedby

productandserviceattributes

CharacterizationoftheCRsaccordingtotheirimpactwithinthe commercialoffer;consequentdeterminationoftheirshareinterms ofcustomer(dis)satisfaction

RelevanceindexesR;CS/CDterms

4 Relatingtheinternalsphereofthe processwiththebusinessoutputs

Estimationofthecontributionprovidedbyprocessphasein fulfillingtheCRs

Correlationcoefficientskij

5 Measuringthephasesexpenditures Extentofemployedresources,emergingharmfuleffects,auxiliary functions,costsandtimenecessarytocarryouttheprocessphases

Phasesresourceindexes

Table4

OutputsofthePVAprocedureandstepsinwhichtheycomeout;thecolumnOutcomesillustratesthekindofobtainedresultsintermsofthecharacterizationofthetreated industrialprocess.

PVAstep Tasktobecarriedout Outcomes Procedureoutputs 6 Measuringtheprocessoutputsfrom

customerviewpoint

Benefitsdeliveredbyeachphaseintermsofavoideddissatisfaction andcustomercontentment

PCSandPCDcoefficients 7 Comparingthedeliveredbenefitsand

theinternalexpenditures

Ratiobetweenthetermsexpressingsatisfactionandthephase consumedresources

VNandVEcoefficients 8 Summarizingtheresults Comparisonofphasesvalue,highlightingofthebottlenecks ValueAssessmentChartgraph

expressed with R, meant as the relative importance of the

customerrequirementswithinthebundleofbenefitsprovidedby

thebusinessprocess.Suchcoefficientsareexpressedwithnatural

numbers through a Likert-type scale, ranging from 1 to 5 in

previous PVA applications. Both the Kano categories and the

relevanceindexesareestablishedthroughcustomers’interviews

orstatedbybusinessexperts,whereasopinionsoftheclientele

areunavailableorconsidereduntrustworthyintheperspectiveof

reengineeringtasks.

The introduction of all Kano categories and importance

coefficientsallowstoestablishtheextentofcustomersatisfaction

CS(dissatisfaction CD)that is generated(prevented). Attractive

andOne-dimensionalattributescontribute(totheextentoftheir

relevanceR)totheCSterm.Must-beandOne-dimensionalfeatures

contribute(stillaccordingtotheirrelevanceR)totheCDterm.The

followingformulasclarify theway CSandCD oftheithCRare

calculated: CSi¼ aiþoi ð Þ P Rk ; (1) CDi¼ miþ oi ð Þ P Rk ; (2)

whereas oi, ai and mi hold Ri if the customer requirement is

classifiedasOne-dimensional,AttractiveorMust-be,respectively,

0 otherwise.The denominator stands for thesum of relevance

indexesascribedtoalltheattributes.

The following phase consistsin identifying, on thebasis of

business experts’ evaluation, the correlation coefficients kij

(0kij1)that linkthephasejtoeachattribute iin termsof

theaccountedrelativecontributionofeachprocessstagetofulfil

thecustomerrequirements(seeFig.1forabettercomprehension).

Theseparametersallowestablishingaquantitativelinkamongthe

processstepsandthearisenbenefits.Itfollowsthateachphaseis

assessed withrespectto itsrole in both providing unexpected

benefitsperceivedinthemarketplace(PhaseCustomer

Satisfac-tion,PCS)andavoidinguserdiscontent(PhaseCustomer

Dissatis-faction, PCD), according to the following mathematical

expressions: PCSj¼ X i kijCSi; (3) PCDj¼ X i kijCDi: (4)

Theextentofemployedresources,emerging harmfuleffects,

costsandtimenecessarytocarryouttheprocessphases(globally

indicatedasRESjfortheshareofthejthphase)arecomparedwith

the terms expressing the customer satisfaction (PCS and PCD).

Thus, the phases are estimated in terms of their capability to

providebothbasicqualityandunexpectedfeatures,accordingto

their consumption, throughout the terms Value for Exciting

requirements (VEj) and the Value for Needed requirements

(VNj),asfollows: VEj¼ PCSj RESj ; (5) VNj¼ PCDj RESj (6)

The outcomes of the analysis (normalized VEj and VNj) are

summarizedinadiagram,namelyValueAssessmentChart(VAC),

standinginaschemewithfourquadrantsthatrepresentdifferent

performance areasfor processphasesaccording toVEjand VNj

indexes(Fig.2).Morespecifically,thequadrantsrepresent:

 AreaofLowValue(lowVEjandVNj,e.g.Phases2and5inFig.2):

theemployedresourcesdonotguaranteeanadequateproduct/

service appreciation level and their extent is excessive with

respecttotheshareofconsumerdissatisfactiontheyarecapable

toavoid.Consequently,reengineeringactionsshouldmassively

regard thephasesfalling insucha quadrant. Said phasesare

demandedto delivernovel functionsand shouldberadically

restructured in ordertodrop their resourcesconsumption or

eventrimmedandsubstitutedbyexistingprocesssegments.

 AreaofBasicValue(lowVEjandhighVNj,e.g.Phase4inFig.2):

the employed resources do not provide perceivable product/

serviceunexpectedbenefits,but theyare wellspentto avoid

consumer dissatisfaction; typically, thephases falling in this

quadrantareorientedtofulfilthefundamentalattributesanddo

notnecessarilyrequireinvestments.

 AreaofExciting Value (high VEjand low VNj,e.g. Phase1 in

Fig.2):inthiscase,theemployedresourcesplayanevidentrole

toproduceagoodproduct/serviceappreciationlevel,buttheydo

not contribute toavoid consumer dissatisfaction; the phases

fallinginthisAreaareworthofinvestmentsinordertomaximize

theirgeneratedbenefits;theirsuccessisakeytolettheproduct/

servicedifferentiatefromthecompetitors.

 AreaofHighValue(highVEjandVNj,e.g.Phase3inFig.2):this

quadrant is characterized by phases capable to provide well

perceivablebenefits,stillmaintaininganextremeefficiencyfor

fulfillingbasicneeds;thephasesbelongingtothisAreaaretobe

safeguarded.

3.2. Deficienciesoftheoriginalmethodology

ThefirstindustrialapplicationsofthePVAshoweditscapability

toorientatedecisionsconcerningBPR.However,asassessedbythe

developers of the methodology, consistent limitations of the

methodology concern subjective evaluations about its input

parameters[54].Theoutputs,resultingbytheapplicationofthe

procedure,are likely to suffer fromimprecision and variability

whicharecurrentlynotmonitored.

The uncertainties may regard the classification and the

relevanceofcustomerrequirements,theratesofprocessphases

infulfillingtheaforementionedattributes,theamountofresources

spentbyeach industrial activity whennot directly measurable

throughmonetaryexpenditures.Besides,thelistofphasesandthe

attributesthroughwhichthebusinessprocessisschematizedcan

differaccordingtosingleanalysts.

Giventheaboveconsiderations,thedifferentestimationscan

beimputedtotheabsenceofadeterministicmodelfortheanalysis

of the system and by diverging evaluations of the process

parameters according to experts’ role (e.g. by considering the

perspective of account executives and industrial production

managers). Consequently, within the inputs of PVA model,

epistemic uncertainty (i.e. related to the lack of knowledge or

causedbymeasurementerrors)issupposedtobemoreimpacting

thanaleatoricuncertainty(i.e.provokedbythevariabilityofthe

involvedparameters),althoughthelatterisnotnegligible.

3.3. Expectedenhancementsandmethodologicalrequirements

Themethodologicalobjectiveofthepresentpaperistherefore

toimprovethedecisionsupportprovidedbythePVA,throughthe

adoptionofanupgradedmodelcapabletomanageuncertainties.

Table 5 summarizes the main differences between the original

algorithm and the new proposal, as well as the expected

achievementsintheindustrialpracticethroughtheconsideration

ofuncertainty.

Ashighlightedin Table5,thenewsystemis meanttowork

whenmultipleusers analyseanindustrial processby meansof

PVA,thusrevealingtheextentandthenatureofdivergences.Since

subjectivityprimarilyaffectstheinputsoftheoriginal

methodol-ogy (as already recalled in Section 3.2), the variability of said

parameters hastobetakeninto account.The variabilityof the

outputsoriginatedfromapluralityofPVAapplicationscouldbe

easily managed by means of descriptive statistics, but such a

strategyis unsuitableinlightof likelyhighsensitivityfromthe

inputs. Eventually, a particular attention has to be dedicated

towardsemployingtechniquestailoredtodealwiththekind of

uncertaintythatcharacterizesthemethodology.

4. Methodologicalapproachforthedevelopmentofan

enhancedPVA

Onthebasisoftheabovediscussion,thedevelopmentofthe

PVAhastoincludetheexploitationofdatagatheredfrommultiple

sources, thus taking into account process uncertainties and

estimatingtheirimpactwithreferencetotheendresults.

Table5

Maindifferencesbetweenthenewproposalandthepreviousreference.

Feature OriginalPVA UpgradeddesignofthePVA

Wayofworking Post-processingofprocessanalysisperformedbyasingleindustrial expert

Post-processingofprocessanalysesperformedbymultiple subjects

Providedinformation Potentiallyunreliablepictureofprocessbottlenecksandstrengths Pictureofthemostprobableprocessbottlenecksandstrengths Use DeterminationofthemosturgentBPRmeasurestobeundertaken DeterminationofthemosturgentBPRmeasurestobe

undertaken;considerationoftheriskassociatedwith redesigningthestructureofindustrialprocesses Fig.2.Illustrativeexampleofthefour-quadrantschemenamedValueAssessmentChart.

SincePVA followsan algorithmiclogic,transforming variables

initiallyintroducedintocharacteristicvaluesfortheprocessphases,

uncertaintyaboutsaidinputshastobeconsideredinordertoprovide

afullspectrumofpossibleoutputs.Onthesepremises,MonteCarlo

simulation method represents an acknowledged opportunity for

dealingwiththeuncertaintyofinputswithincomplexmathematical

models [55,56]. Monte Carlo method is a widespread technique

tailored to support decisions, due to its capability to generate

scenarios according to many varying and uncertain inputs. Its

employmentiswidelywitnessedinnumerousdomains,including

engineering[57],productdevelopment[58],business[59]andproject

management[60].Withaparticularfocusonengineering

applica-tions,themethodistailored,asassessedbyKreinovichetal.[61],to

dealwithbothepistemicandaleatoricuncertainty.

At this stage, the objective is therefore replicating a large

quantityofinputsbasedonthedatathathavebeenintroducedbya

limitednumberofindustrialexpertsapplyingthePVAonthesame

businessprocess.TheapplicationofMonteCarlomethodtoPVA,

byinvolvingthewholerangeofpotentialvaluesforeachinput(i.e.

numberof phasesandcustomerrequirements,Kanocategories,

relevanceextents,correlationindexes kij,amountofresources),

wouldresultmerelyinasensitivityanalysismeasuringtheoverall

potential impact on the outputs, thus out of the scope of the

presentresearch.Willingtoconstrainthesimulationwithrespect

tospecificinputs,thecommonapproachistoreplicateeachinput

variable according to an attributed probability distribution

function,whichbestfitspreviouslygathereddata.

AnexplanationisgiveninthefollowingSubsectionsaboutthe

choicesperformedtocarryoutthesimulation.

4.1. Simulationproblems

4.1.1. Problemswithdivergingsetsofphasesandcustomer

requirements

Itisexpectedthatexpertsofagivenbusinessactivitysharea

commonvisionabouthowtheindustrialprocessisorganizedand

whichoutputsareofferedtothecustomerintermsofproductand

service attributes. Propermodelling techniques, likeIDEF0, are

appraisedtoprovideaschematicpictureoftheprocessandthusto

rapidlyfindaconsensusaboutthesystemboundaries,thephases

andtheiroutcomes.

However,discordancesaboutthedeterminationofthephases

and thelistof customerrequirements could still emerge.Such

mismatchesdetermineconsistentproblemsinthesimulationtask

that would be better performed starting from homogeneous

systemsofphasesandoutputsvaluableforthecustomer.Inorder

to address such an issue, the schema of an overall industrial

processisfavourablyrepresentedbyincludingallthephasesand

requirementsthathavebeenindividuatedbytheanalysts.With

referencetothemissingphasesor attributesofeachindividual

analysis,thefollowingstrategycanbefollowed:

 allthecorrelationindexes andresourceratios concerning the

phasesthatarenotrepresentedbyanalystswillbeassignedthe

nullvalue; in this way suchphases do not play any role in

deliveringcustomervalue,nordotheyimpactprocess

expen-ditures;

 therelevanceextentsofthecustomerrequirementsthatarenot

judgedimpacting byanalystswillassumethe nullvalue,thus

providingnocontributioninthedeterminationofcustomervalue.

4.1.2. Problemswiththesimulationofnominalvariables(Kano

categories)

WithinPVA,Kanocategoriesinputsarenominalvariables,for

which Monte Carlo method is not applicable. Conversely, the

simulationof categoricalvariables isusually addressedthrough

resamplingtechniques.Inparticular,thebootstrappingmethodis

capabletoreplacesophisticatedmathematicalproceduresthanks

to thegrowing computationalpower ofcalculators [62]and is

tailoredforexperimentscharacterizedbysmallpilotsamples[63].

The major shortcoming of such a technique stands in the

impossibilitytoreplicate nominalvariables that havenot been

indicatedbyanyanalyst[64].

Inordertoovercometheproblem,complexstatisticaltoolscan

improveresamplingapplicationsotherwisebiasedbytheabsence

of some, although scarcely likely, nominal values [65]. This

approach does not fit however thescope of buildingan easily

usablesystem.Asaconsequence,theauthorsoptedtosimulateCS/

CD variables, which containthe information provided by Kano

categories,i.e. thekindof customerperceptionabouta specific

productrequirement.

4.2. Simulationchoices

4.2.1. SimulatingCS/CDindexes

AsclarifiedinSection4.1.2,itisnecessarytosimulateCS/CD

indexes for each surveyed productattribute. CS/CD coefficients

range, by definition,in the [01] interval and can be therefore

supposed to follow a beta distribution,for which Monte Carlo

simulationisapplicable[66,67].Sincetwoshapeparametersare

requiredfortheutilizationofthebetaprobabilitydensityfunction,

thesecoefficientscanbecalculatedbyexploitingthemeanandthe

variance of thesample data (i.e.CS/CD indexesemerging from

individualPVAanalyses),asin[68].

The determination of thescale parameters allows therefore

drawing,foreachcustomerrequirement,anarrayofsimulations

followingasuitablebetadistribution.Thesizeofthearraydepends

onthepreviouslyestablishednumberofsimulationsthatwillbe

indicatedwithnsiminthefollowings.

4.2.2. Simulatingcorrelationindexes

The kij coefficients that express the role of the phases in

fulfilling each customer requirement are positive uncertain

variables with a fixed sum, i.e. 1. According to the authors’

experience, the analystsof industrial processes employingPVA

individuate for each product attribute one influential phase

playingamajorroleforthefulfilmentoftheattributeitself.The

influenceoftheresidualphasesisestablishedwithrespecttothe

key one.Byobservingthis characteristicbehaviour,theauthors

havechosenaspecificsimulationstrategyforkijvariables.

A single reference phase is individuated for each product

attribute, by selecting theone with the maximum average kij

withintheprocessanalyses.Subsequently,foreachanalysis,the

procedurerequirestocomputetheratios(indicatedwithk0

ijin

the followings) between each individual kij and the values

concerningthekeyphase.Suchratiosare,bydefinition,positive

or null variables. Their mean and variance have to be

subsequently calculated. k0

ij ratios showing non-nullvariance

(atleastthereferencephasehasnotthisfeature)aresupposed

to follow a gamma distribution (being they positive or null),

which allows the application of the Monte Carlo simulation

method [66]. The definitionof the gamma probabilitydensity

functionrequirestheknowledgeoftwoparameters(shapeand

scale), that can be deducted by mean and variance also for

simulation purposes, as in [69]. Hence, the varying k0

ij are

simulated nsim times, while the constant ones are simply

repeated in the same quantity. Subsequently, arrays of kij

simulated coefficientsare determinedby turningtheemerged

proportionsintosharessummingto1.Forany givencustomer

requirementandwithreferencetoaspecificsimulation,suchan

outcomecanbetriviallyobtainedbydividingeachsetofratios

4.2.3. Simulatingresourcesshares

As seen for kij coefficients, resource shares are uncertain

variableswhosesumequalsto1.Ifanyprobabilitydistributionis

establishedforeachshare,thepresenceofthefixedsummakesthe

simulation problem over-constrained. Withrespect toresource

shares, no roundabout strategy can be applied, since analysts

employsubstantiallydifferentcriteriatodeterminetheimpactof

phasesonoverallprocessexpendituresandoperationtimes.

Suchakindofproblemhasbeenhoweverfacedintheliterature,

byexploitingconditionalprobabilitytheory[70,71].Byassuminga

normal distribution for the treated variables, it is possible to

exploitavailablesimulationstrategiessuitableforimplementation

throughspecificprogramminglanguages.Theauthorshavethen

partiallyadoptedthelogic andthecommandssuggested byan

Internet resource supporting the development of scripts in

MATLAB1_{.Themeansand thevariancesofthesample dataare}

required to carry out thesimulation, giving rise to nsim-sized

arrays of resource ratios regarding each individuated process

phase.

4.3. Stepwiseguidedmethodologyandsoftwareapplication

This Subsection provides a stepwise guide to apply the

proposedmethodology,asillustratedinFig.3.Inadditiontothe

sequenceof the steps, thefigure highlights theintervals tobe

considered for carrying out each operation. The illustration is

organizedin three sub-diagrams (delimited by rectangles with

dottedlines),standingforthe‘‘Gatheringofindividualanalyses’’,

the‘‘Elaboration ofobtained data’’and the‘‘Simulationprocess

leadingtodisplaythefinalresults’’;hereinafter,theyaredescribed

indetail.

4.3.1. Gatheringofindividualanalyses

Withrespecttotheactivitiesincludedinthissub-diagram,the

user has to collect the complete list of individuated business

process phases and customer requirements, thus building an

overallschemaofthesystemunderinvestigation.Missingdataasa

result of disregarded phases and product characteristics are

introduced,asdescribedinSection4.1.1.

4.3.2. Elaborationofobtaineddata

Oncetheindividualanalysesaregathered,theobtaineddata

shouldbeorganizedasfollows(thenumbersofthefollowingitems

correspond tothose reported in Fig.3 on the arrowsfeaturing

differentstreamsofdataelaboration):

(1) CS/CD indexes are calculated for each listed customer

requirement and each individual PVA analysis through the

expressions(1)and(2);theirmeanandvariancearecalculated

for each customer requirement, so as to determine shape

parametersforfittingabetadistribution;

(2) anarrayoftheassignedcorrelationindexeskijhastobebuilt

foreachlistedcustomerrequirement,processphaseandPVA

analysis;thedataarethendividedforthevalueofkijshowing

the maximum mean for each customer requirement, so

determiningk0

ijindexes;meansandvariancesofk0ijvariables

arecalculatedandconsequently shapeandscaleparameters

arecomputedforfittingagammadistribution;

(3) anarrayoftheattributedresourceratioshastobearrangedfor

each listed process phase; their means and variances are

subsequentlycalculated.

Inordertoruntheoperationsrecalledinthethirdsub-diagram,

the number of the steps of the simulation hasto be planned

(usually somethousands),by takinginto account theexpected

reliabilityoftheMonteCarlomethodoutcomes[72].Then,onthe

basisof theinputs resulting fromthesecond sub-diagram,the

simulation is performed accordingtotheattributedprobability

densityfunctions,asdescribedinSection4.2.Thisallowsdrawing

nsim-sizedarraysofresourceratios,CS/CDindexesandkijshares

(whicharecalculatedafterthesimulationofk0

ijvalues).Thedata

arethen usedfor simulating nsim PVAanalyses, leading tothe

samequantityofVE/VNpairs,bymeansofformulas(3)–(6).The

emerging data are used to assess uncertainties about the

performancesofthephasesin theexaminedindustrial process.

Suchuncertainties canbegraphicallyrepresentedinamodified

VAC diagram, which replaces single points symbolizing phases

performanceswithsetsofthemostlikelyVE/VNvalues.

4.3.3. Simulationprocessleadingtodisplaythefinalresults

Thissub-diagramrepresentstheportionofthealgorithmthat

can beautomatically executedby exploiting thebuilt MATLAB

script,freelydownloadableasanopen-sourcewebresource2_.Such

a computer application helps carrying out the methodology

especially in those parts that require greater computational

efforts.Theresidual stepscanbeeasilyperformedmanually or

throughdiffusedsoftwaretools,suchasspreadsheets,sincethey

require just trivial mathematical operations (multiplications,

divisions, determination of means and variances). Then, data

obtainedintheinitialsteps(thefirsttwosub-diagramsofFig.3)

havetobeproperlyintroducedintheMATLABscriptinorderto

correctlyrunthesimulation,asspecifiedinthedescriptionofthe

routine at the indicated webpage. The software application

includesamainpart,devotedtocarryoutthesimulationandto

computetheperformanceofthephases,andaconclusiveblockto

extrapolatetheVACdiagram.

Such a moduledisplaysprocessperformance throughactual

simulated data, creating for each phase a broken line that

delimitates the most populated area of the graph in terms of

VE/VNvalues.

5. Descriptionoftheexperimentsanddiscussionoftheresults

The performed tests have been designed to check the

applicabilityofthemethodologyandofthesoftwaretool,aswell

asthereliabilityoftheoutputs.

Afirstexperimenthasinvolvedpeoplewithpoorexperiencein

theindustrialdomain(pharmaceutics)oftheexaminedbusiness

process.Theprovidedinformationwashoweversufficientforeach

experimentertosketchananalysisthroughthePVA.Saidprocess

regardswellestablishedmanufacturingpractices and

technolo-gies,whoseevolutionsareknown.Theemergingresults,interms

ofdecisionsdictatedbytheVACdiagramandphasesuncertainties,

werethereforecomparedwiththerealobservedtransformations

oftheinvestigatedprocess.

Peoplewithgreaterknowledgeoftheindustrythenconducted

asecondexperiment.Theanalysedindustrialprocessregardsthe

currentactivitiesofaSMEproducingwomen’sshoes,involvedina

research project together with the Institution of some of the

authors. The sample of testers included some member of the

managementof theshoefactoryand volunteerstudentswitha

goodlevelofknowledgeabouttheanalysed industryandfirm’s

practices. The outcomes of the application of the proposed

methodologywere then illustratedto thefactory’sdirectionin

order to elaborate more conscious decisions about process

reengineering. 1 www.mathworks.it/matlabcentral/newsreader/view_thread/304141. 2 http://www.mathworks.com/matlabcentral/fileexchange/44594-pva-simula-tion-with-monte-carlo.

Fig.3.Overviewofthesimulationprocessconcerningthewholesampleofcustomerrequirementsandprocesssegments;full-linecirclesindicatetheintervalinwhichto containthedataemployedintheconnectedmethodologicalsteps;thebottomdotted-linequadrantindividuatesthepartofthealgorithmsupportedbythedeveloped MATLABscript.

Ultimately,thetwotestsaresupposedtoprovide

complemen-taryresults.Whereastheformeraimsatcheckingtheapplicability

ofthemethodologyandthereliabilityoftheoutputs,thelattercan

be exploited to evaluate its capability to support decisions in

industrialenvironments.Thetestsaredescribedinthefollowing

subsections.

5.1. Pilotexperiment

TheSectionillustratesanexperimentabouttheapplicationof

the original PVA, carried out by a sample of convenience

constitutedby27volunteerMSEngineeringstudents,attending

the course ‘‘Methods and Tools for SystematicInnovation’’ at

PolitecnicodiMilano(Italy).Thestudentsperformedthetesting

activitybycompilingatailoredspreadsheet,which

automatical-ly computes the main outcomes and graphically displays the

outputs of the methodology throughout the original VAC

diagram.

The case study, extracted from a real industrial project,

regardsawell-established processfor treatingpharmaceutical

powdersinordertoenhancethemanufacturabilityoftablets,i.e.

high-speed granulation. More in detail, the objective of the

analysed process stands in the transformation of Active

Pharmaceutical Ingredients (APIs) and excipients, commonly

deliveredin a powder state, into grains. Such grains have to

showagoodlevelofflowabilitytobeeasilycompressedforthe

manufacturing of tablets. The technique consists in a prior

mixingofwater,APIandexcipientpowdersinordertoobtaina

doughycompound. Thesubsequent phaseconsistsinchipping

the dough (a sort of extrusion) into filaments. Such formed

productsare,inturn,driedbeforebeingsubmittedtoafurther

crackingsotoobtain structureswiththerequestedsizeofthe

grains.Theobtainedgrainsare subsequentlysifted toselecta

sufficiently homogeneous output showing the required

char-acteristicsofflowability.

Theinvolvedprocessischaracterizedbyconsistentinformation

about its effective evolution, obtained throughout experts

involvement in the research partially described in Becattini

et al. [73]. More specifically, actual process developments,

andhenceexpectedindicationsprovidedbymethodologystand

in:

 thereductionofchannelledresourcesforthemixingphase;

 the removal of the phasesdevoted to reduce thesize of the

pharmaceutical material (as observed in the fluidized-bed

technology)ortheirintegrationwithotherprocessphases(like

itisperformedwiththespry-dryingtechnology);

 the key attention paid to the drying phase or to alternative

activities aimed at maintaining a well-defined extent of

humidity;

 atechnologicalchangeforthesiftingprocess,withtheobjective

ofstronglyreducingtheemployedresources.

Thedescriptionoftheprocessprovidedtotheexperimenters

reportstheavailableinformationaboutthegranulation

technolo-gy,soastoallowtheschematizationoftheindustrialsystemand

extracttheknowledgerelevantforPVAtasks.

On the basis of the process description, all the 27 testers

individuatedthesameprocessphases,consistinginthemixing,

doughextrusion,drying,filamentschippingandsiftingoperations.

Each studentdescribed theoutputs of thetechnical system by

meansofanumberofCustomerRequirementsrangingfrom6to

10. The whole sample of CRs, that takes into account all the

individuated features mentioned in the complete analysis,

includes 10 items, further on named CR1 to CR10: dosage

homogeneity, density and porosity, flowability, size, relative

humidity, reduced volatility and contamination, mechanical

characteristics, hardness, smoothness and aesthetic properties,

colourhomogeneity.

TheapplicationofthemethodologyexploitedadifferentVAC

representationtoavoidoverlappinglines.Thenewillustration(see

Fig.4)makesuseofParzenwindowing[74],awidespread

non-parametricprobabilitydensitydistribution.

Fig.4.RepresentationofuncertaintiesthroughthemodifiedVACdiagramwithreferencetothegranulationprocessofpharmaceuticalpowders:Parzenwindowingis employedforthescopesoftherepresentation.

Inordertoperformsuchatypeofrepresentation,thefinalblock

oftheMATLABscriptwassubstitutedbyanalternativeroutine3,

whosedraftingbenefittedfromanavailableInternetresource4_.In

detail,thestandarddeviationofthesampleswasexploitedforthe

scopes of building Parzen windows, according to the common

normal distribution approximation. Parzen windowing is

sup-posedtoclearlyidentifythemostproperareasoftheVACdiagram

concerningeachphaseforthespecificcasestudy.

Withrespecttothepresentexperiment,theemergingresults

schematizedinFig.4,showthat:

 themainprocessvaluebottleneck,consistinginthesiftingphase,

isindividuatedwithoutrelevantuncertainty;

 alsothedoughextrusionphaseissupposedtopoorlycontribute

tocustomersatisfactionaccordingtotheconsumedresources,

sinceanotnegligibleamountofVE/VNpairsisincludedinthe

LowValueareaoftheVAC;

 theoperationsregardingthedryingandthefilamentschipping

phases result the most value adding; while the former is

particularly oriented towards the fulfilment of less expected

characteristics,thelatterismostlyaddressedatdeliveringthe

basicpropertiesofthegranulationprocess;

 uncertaintiesaboutthemixingphasedonotallowtosuggestany

suitabledirectionforundertakingBPRtasks.

Althoughaffected by uncertainty, theresults provideuseful

information for decision making. According to the general

indicationsprovidedbythePVAmethod,thephasesthatrepresent

value bottlenecks should be submitted to the most severe

transformations within the technological development of the

granulationprocess.

The poor value emerging by the sifting phase suggests

technical changes with respect to the grain separation; as a

matterof fact,less consuming pneumatic sieving are used in

pharmaceuticalindustry,graduallyreplacingmechanicaldevices.

Withrespecttothephaseconcerningthedoughextrusion,also

showinglimitedcontributiontocustomervalue,majorchanges

shouldbeexpected.Aswell,themost diffusedalternativewet

granulationtechnique,i.e.fluidizedbed,recurstoasinglephase

for determiningthe rightsize of thegrains, thusavoidingthe

preliminary volume reduction of the pharmaceutical mix.

Furthermore,thekeyroleplayedbythedryingphaseisremarked

bytheoutcomes.

Besides,theurgencyofloweringtherequiredresourcesforthe

mixingphaseisnotidentified,duetohighuncertainty.Inaddition,

thefilamentschippingdoesnotemergeasaphaseexpectedtobe

overcome,sinceitshowsgoodperformances.

5.2. Industrialapplication

Thesecondexperimentdealswithanindustrialcasestudyfrom

thefootwearsectorandshowsthecapabilityofthemethodology

toorientatedecisionsinindustrial contexts.Moreindetail, the

proposed method has been applied to analyse the design and

manufacturingofshoesforafactorythathasparticipatedtothe

project‘‘ICT4SHOES’’,fundedbyTuscanyRegion,Italy.Theproject

aimsatintroducingnewICTsolutionsfortheproductionandthe

managementofbusinessprocessesinthefootwearindustry.The

accomplishmentofthetaskfirstlyrequiresadeepknowledgeof

industrial processes in order to generate tailored computer

supports. The proposed methodology has been considered a

reference for analysing processes and determining the main

bottlenecks, hence individuating the firm activities requiring

majorredesign.

Themainproductionofthefirmconsistsinsummerwomen’s

fashion shoes, characterized by remarkable lightness and

flexibility. Experts who analysefashion trends firstly planthe

style of the collections. The design sketches have to be

subsequently transformed into physical prototypes andtested

in order to check if the shoes satisfy aesthetics and comfort

expectations. Once the prototypeshave beenrefined after the

test,salesstartbyparticipatingtosectorfairsandentrustingthe

commercial promotionof theitemsto salesmen, who operate

worldwide. New customers can even purchase shoes and

perform ordersthrougha webplatform. Assalesprogress,the

management of the factoryand the commercialunit acquires

orders.Onthisbasis,theyplanthemanufacturingoftheorders,

supervisethestockofmaterials(mainlytheleather)and

semi-finished products (such as heels, soles and accessories). The

organization of the production involves also the choice of

contracting firms that are in charge of developing the initial

modelstoallowthecreationofvarioussizes,manufacturingthe

dies, cutting the leather and making shoe uppers by sewing

leatherparts.Thewarehousingunitofthefactoryisinchargeof

receivingandsendingtotheotherpartiesallthematerials,

semi-finishedproducts andworking instruments. Onceshoe uppers

andalltheremainingcomponentsareavailable,theshoefactory

initiates the assembling phase or supervises this activity if

carried outby thirdparties. The manufacturedshoes arethen

finished,checkedandpackaged,soastoallowtheshipmentof

theordereditemsintherequestedquantityandtypology.

The authorsschematized a model of the business process,

includingphasesandfulfilledcustomerrequirements.Themodel

wasinspectedandmodifiedbythefirm’smanagementuptothe

determinationofaframeworkcomprising12processsegments

and17productattributes.Additionalinformationwasacquired

in ordertoachieve sufficientknowledgeforapplyingthePVA.

The debate with the production manager led to a reference

versionofPVAfortheinvestigatedindustrialprocess.Overall,13

volunteer MS Engineering students, attending the course

‘‘Product Development and Engineering Design’’ at University

of Florence (Italy), took part of the experiment. They were

introduced to the logic of the PVA, taught about the

funda-mentalsoffootwearindustryandputintocontactwiththeshoe

factoryinordertoobtainanyinformationtheyjudgedrelevantto

correctlyperformtheanalysisofthegivenprocess.Thestudents

wereurgedtoacquire independentlyfurther information,thus

providing added value for the scope of the analysis of the

industrialprocess.Attheendoftheprocedure,thestudentsand

3othermembersoftheenterprise’smanagementwereaskedto

modifythereferenceframeworkaccordingtotheirviewpointon

the marketof theshoes,theprocess andits mechanismsthat

enabletheaccomplishmentoftheproductattributes.

The whole sample of 16 examinations through the PVA

performedbyindividualswithanotnegligibleknowledgeinthe

fieldrepresentedthestartingpointforcarryingoutthesimulation.

Thedatawerethengroupedandorganizedinordertoexecutethe

simulationwiththeproposedMATLABtool5_.

Given the great quantity of phases that characterize the

industrialapplication,itwasevaluatedthataseparategraphical

output for each processsegment was preferable. The standard

representationwithbrokenlineswaskept,butthepossibilityto

introduce different thresholds was introduced. The strategy

3_{Thescriptcanbedownloadedfromthewebpage:http://www.mathworks.com/}

matlabcentral/fileexchange/44595-parzen-representation-of-pva-simulations.The partofthemainblocktobesubstitutedisintroducedbythedisclaimer‘‘%THE FOLLOWINGMODULEISAIMEDATBUILDINGGRAPHICS’’.

4

http://stackoverflow.com/questions/9134014/contour-plot-coloured-by-clus-tering-of-points-matlab.

5

Thedataemployedtoperformthesimulationareavailableinthefirstcomment concerningthefileexchangepageofMATLAB,wherethescriptisreported.

exploiteda secondalternative MATLABblock6_{, substituting the}

finalpartofthemainscript.

The diagramshownin Fig.5 isa resultofthecomputerized

procedureregardingthephaseentrustedtodeterminethestyleof

collections.Althoughtheextentofuncertaintiesisconsiderablefor

thisphase, thedefinition ofthestyle canbe considered a well

performingactivity(inthecontextoftheanalyzedprocess),being

itscorrespondingVE/VNpairsmajorlypositionedintheHighValue

area of the VAC. Other phases characterized by significant

uncertaintiesfaceasituationforwhichnoquadrant oftheVAC

ispredominantanddecisionscanbehardlytaken,e.g.testingof

prototypes,whichisschematizedinFig.6.

Forthescopeoftacklingreengineeringinitiativesintheshoe

factory,thecoreof theanalysisstands inthe individuationof

valuebottlenecks.Beforetheapplicationofthemethodology,the

enterprise had already individuated the need to update the

technologiesemployedforwarehousingactivities.Fig.7,showing

thediagramrelatedtosuchaphase,partiallyconfirmsthischoice,

beingVE/VNpairsconcerningwarehousingdiffusedlyplacedin

the Low Value area. Anyway, other quadrants are rather

populated,especiallyBasicValuearea.Ontheotherhand,several

manufacturing phases clearly represent process bottlenecks,

since,althoughuncertaintiesarepresent,veryfewVE/VNpairs

lie outside of the Low Value area. Figs. 8–10 schematize the

performances of leather cutting, leather sewing and shoes

assembling,respectively.Itcanbeadditionallyunderlinedhow

representativeareaswiththresholdssetat75% and95%ofthe

wholesimulationofVE/VNvalueswidelyoverlap,especiallyin

Figs.8and9.Therefore,themeasuresofthephasesperformance

are highly concentrated and this lessens the risks about the

decisionstobeundertaken.

Theapplicationofthemethodologyconvincedtheshoefactory

to rethink its plans for process reengineering, considering to

includealsomanufacturingactivitieswithinthebundleoftasksto

beredesignedinordertoenhancefirm’sbusinessoutcomes.

6. Discussionandconclusions

Despite different methodological options should be tested

(such as systems for BPR evolution, process-oriented QFD, as

discussedattheendofSection2)inordertocomparetheefficacy

of alternative approaches, the candidate module for a DSS

reported in the paper has demonstrated its capabilities with

respect to the objective of the present research. Indeed, the

illustratedmethodology,consistinginaradicalreworkofthePVA

[10], supports the individuation of the main deficiencies

pertainingtheinvestigatedindustrialprocesstowardsthegoal

of delineating the reengineering priorities and consequently

applyingthemostbeneficialBPRtools.Withrespecttotheposed

requirements,themodulepursuesthedoublegoaloftakinginto

accountthecustomersphereandevaluatingtheriskassociated

withthedecisions tobeundertaken, accordingtothe levelof

disagreementamongthedecisionmakers.

Moreindetail,theillustratedsystemworkslikethoseDSSsthat

integrateinformationpertainingtheenduserdomain.Indeed,it

usessaidinformationtobuildquantitativevaluemetricsandtakes

intoconsiderationalsotheuncertaintiesconcerningsuchissuesin

ordertostrengthenthereliabilityoftheoutputs.Inthedeveloped

model,theassessmentofuncertaintiesimpacthasbeen

accom-plishedbyintegratingspecificsimulationtoolswithintheoriginal

methodology.Theproposedapproach issuitable for supporting

decisiontasksinsituationscharacterizedbyanyofthefollowing

circumstances: superficial information concerning customer

opinions;urgencyofthedecisionsuchthat itisnotpossibleto

collectexhaustiveorreliabledata;highvariabilityofthecontext;

divergingevaluationsprovidedbysectorexperts.

The original methodology and a novel simulation module,

meanttoallowthehandlingofdiverginginputs,arecharacterized

by theease of being exploited by means of diffused software

applications(suchasspreadsheets).Thetaskisfurthersimplified

byemployingMATLAB,thankstoascriptpublishedontheweb,

which automates the simulation procedure and additional

routinestobuildsuitablegraphicalrepresentations.Consequently,

such tools are suitable also for small-sized firms withlimited

resourcesandcompetencesinstatistics.

Fig.6.Representationofuncertaintiesreferredto thephaseentrustedto test prototypesinthefootwearindustry:contoursofareascompriseatleast75/95%of VE/VNsimulations,asindicatedinthefigureitself.

Fig.5.Representationofuncertaintiesreferredtothephaseentrustedtodetermine thestyleofshoesinthefootwearindustry:contoursofareascompriseatleast75/ 95%ofVE/VNsimulations,asindicatedinthefigureitself.

6

Thescriptcanbedownloadedfromthewebpage:http://www.mathworks.com/ matlabcentral/fileexchange/44596-single-vac-graphs-with-threshold-selection. Thepartofthemainblocktobesubstitutedisintroducedbythedisclaimer‘‘%THE FOLLOWINGMODULEISAIMEDATBUILDINGGRAPHICS’’.

The results emerging from the first verification activity

highlightthatthevaluebottlenecksofabusinessprocesscanbe

identified in cases characterized by diverging evaluations.

Although great uncertainties, low-valued process phases were

individuatedforboththeexperimentsillustratedinSection5.At

the same time, the presented methodology is capable to

individuateprocessactivitiesforwhichreengineeringcouldresult

hazardous.Anyway, improvementsare expectedin light of the

missed identification of advantageous reengineering activities,

which havetaken placein thepharmaceuticsindustry (Section

5.1).Fromthisviewpoint,experimentscarriedoutonlybyexperts

shouldhighlighttheroleplayedbythelimiteddomainknowledge

inaffectingthefinaloutcomesoftheproposedtool.

Moreindetail,theobtainedresultshavehighlightedthatthe

proposedmethod:

 iscapabletoevaluatetheimpactthatuncertaintieshaveonthe

valueindexescharacterizingeachphase;

 allowsestimatingtheuncertaintyoftheprovidedoutputs,hence

the reliability of the consequent reengineering actions that

decisionmakersmightundertake;

 helps the users establishing which aspects of the business

process(ifany)resultmorefuzzyand,thus,whichinformation

elementsrequirefurtherinvestigationsin ordertoreducethe

uncertaintyoftheoutputs.

However,currentlimitationsofthemethodologydonotallowits

applicationinparticularcases,suchasthesituationsthatfollow:

 processesmarkedly characterized by internalor management

phasesthatdonotprovidevalueforcustomerstoaconsiderable

extent, but that arecompulsoryand cannot betrimmed (e.g.

Fig.10.Representationofuncertaintiesreferredtothephaseentrustedtoassemble shoesinthefootwearindustry:contoursofareascompriseatleast75/95%ofVE/VN simulations,asindicatedinthefigureitself.

Fig.7.Representationofuncertaintiesreferredtothephaseentrustedtowarehouse materialsandsemifinishedproductsinthefootwearindustry:contoursofareas compriseatleast75/95%ofVE/VNsimulations,asindicatedinthefigureitself.

Fig. 8. Representation of uncertainties referred to the manufacturing phase entrustedtocutleatherinthefootwearindustry:contoursofareascompriseat least75/95%ofVE/VNsimulations,asindicatedinthefigureitself.

Fig.9.Representationofuncertaintiesreferredtothephaseentrusted tosew leatherinthefootwearindustry:contoursofareascompriseatleast75/95%ofVE/ VNsimulations,asindicatedinthefigureitself.