Volvo IT Belgium has provided real-life datasets for the need of the Third International Business Process Intelligence Challenge, in the form of event logs generated by the VINST system used across the Volvo corporation to support incident and problem handling. Volvo has also pointed 4 aspects of their business operations they would like to be scrutinized. 1) Are the incidents contained within the 1st support line? 2) Is bouncing delegation, a.k.a. ping-pong, a frequent phenomenon? 3) Are employees cheating on VINST by faking inexistent waits from endusers? 4) Are real process instances conform with the process model proposed by Volvo IT? In this paper, we provide answers to all these questions using process mining and social network analysis techniques, and we state the existence of hidden support lines degrading the overall performance of incident handling, little localized ping-pong behavior and wait-user misuse, and various levels of conformity across organizations.
Volvo IT provides IT services according to terms and conditions regulated in
Service Level Agreements (SLAs). The VINST system is used by Volvo IT to
support incidents and problems reported by the IT service users. The primary
goal of the Incident Handling process is to \restore normal service operation as
quickly as possible and by that ensure that the best possible levels of service
quality and availability are maintained" [
Guidelines for the execution of both processes follow from the ITIL
approach [
The data analyzed in this paper come from the VINST system. Volvo IT has pointed the following aspects of request handling as particularly interesting and has asked for in-depth analysis: 1. push to front strategy : this aspect refers to the Volvo IT policy encouraging incident handling mainly in the 1st line STs; every time an incident is forwarded to the 2nd and the 3rd line, it has a negative impact on the process e ciency and the e ciency of the Volvo IT as a whole; 2. ping-pong behavior : this aspect refers to the unwanted pattern in interaction among STs, when a request is repeatedly bounced from one ST to another; 3. wait user sub-status use : this aspect refers to the unwanted mis-use of the \wait-user" activity sub-status available in the VINST system; the overuse of this status indicates the hidden ine ciency of the process; 4. process conformity per organization : this aspect refers to the compliance of the real process instances with an \ideal" designed process model.
Speci c questions were asked by Volvo IT concerning all these four aspects [
In this paper, we present the results of our analysis of the data provided by
Volvo IT, with the help of both process mining [
The provided event log is a record of activities undertaken by Volvo IT
employees during the execution of Incident Handling [
Only slight modi cations of the event log were required in order to start the analysis: rst, data of the \support team" column have been divided into two columns h"support team", \support line number"i, with the values of \support line number" being in the setf\1st", \2nd", \3rd", \2nd-3rd"g; second, the 1st line number was assumed for activities for which no line number was mentioned; third, the format of timestamps has been modi ed to enable their processing by various tools.
During data munging, the following inconsistencies within the datasets has
been identi ed: rst, some statuses and sub-statuses of activities not speci ed
in the dataset [
The Incident Handling event log captures 65.533 events generated during the execution of 7.554 process instances. The events were recorded in a time period from 31.10.2010, 15:59:42 to 23.05.2012, 00:22:25. 81% of all the events have been recorded from April, 16th to May, 19th 2012. Activities are associated with 705 di erent products. Products associated with the highest numbers of process instances are products 424 (882 process instances, 11.6%), 660 (484, 6%), 253 (226, 2.9%), 383 (205, 5%).
There are 13 activity types in the Incident Handling log: \Accepted/In Progress" (46.14% of all the activities), \Queued/Awaiting Assignment" (17.62%), \Completed/Resolved" (9.33%), \Completed/Closed" (8.72%), \Accepted/Wait - User" (6.43%), \Accepted/Assigned" (4.92%), \Completed/In Call" (3.11%), \Accepted/Wait" (2.34%), \Accepted/Wait - Implementation" (0.75%), \Accepted/Wait - Vendor" (0.48%), \Accepted/Wait - Customer" (0.15%), \Unmatched/Unmatched" (0.01%), \Completed/Cancelled" (0.001%).
The majority of Handle Incident process instances starts with an \Accept/In Progress" (84.35%) or \Queued/Awaiting Assignment" (15.3%) activity. Four activities never start the Incident Handling process: \Assigned/Wait-customer", \Unmatched/Unmatched", \Completed/Closed", \Completed/Cancelled". The majority of process instances is nished by \Completed/Closed" (73.77%) while process instances nished by \Completed/In Call", \Completed/Resolved" and \Completed/Cancelled" account for 26.1% of all the process instances. Eight process instances (0.1%) is still running.
The most frequent behavior is presented in Fig 1. Only the most frequently executed activities and transitions are visible. Numbers assigned to activities and transitions indicate the number of process instances they appeared in.
There are 2.278 variants of Incident Handling process execution. The most frequent variant (1.749 executions, 23.15%) consists of three steps: \Accepted/In Progress", \Accepted/In Progress", and \Completed/In Call". The second most popular process variant (524 executions, 6.94%) consists of four steps: \Accepted/In Progress", \Accepted/In Progress", \Completed/Resolved", \Completed/Closed". Although the number of events varies among variants from 1 to 123, 14% of process instances are performed in less than 10 minutes. The maximal duration of a process instance is 2 years 41 days.
The Incident Handling process is rather unstructured, as illustrated in Fig. 2.
All the activities recorded from April, 16th to May, 19th 2012 are presented in appeared in.
The generalized diagram capturing the most frequent behavior is presented in Fig 1. Only the most frequently executed activities and transitions are visible. Numbers assigned to activities and transitions indicate the number of process instances they 4
Zbigniew Paszkiewicz and Willy Picard
The maximal duration of a process instance is 2 years 41 days.
The IncidFenigtH1Fa.ingTd.lye1p.piTrcoyacpleiscHsaliasHnradantlhdeelerInIunsnststatarnnuccceteurperdo.
procecTseshsiessxieescxuctoeincofnui.rtmioend by Fig 2. In Fig 2a all the log activity recorded from April, 16 to May 19 2012 is presented.
ThereTahreere2is2a7b8igvvaarriiaentytsofotrfanIsnitcioidnesnamtoHnagnadctlievitpiersolceeadsisngextoescpuatgihoentt.i-lTikheedimagorasmt.frequent
Fig. 2a). The wide variety of transitions among activities leads to spaghetti-like one (174D9ueetxoetchueticoonmsp,le2xi3ty,1o5f%th)e tcroannssitiisotnss oamfotnhgeactthivrietiees,stveeprys:ofAtecncienpttheisd/pIanpePr, rogress, diagram. The Incident Handling process is presented in Fig. 2b)as a dotted chart. diagrams present only the most common and frequent behavior. This is to keep the AcceptedE/Ianch Prorwogcroersress,poCnodmstpoleexteadct/lIynonCeaplrl.oceTshseinssteacnocenadndmeoacsht dpootpcuorlarerspvoanrdisant (524 diagrams understandable and readable. Different representation of the process is executiontos,an6a,9ct4iv%it)y. Pharoscefsos uinrstasntecpess:areAscocrteepdtefrdo/mInthePrtoopgraecscso,rdiAngcctoepsttaerdt/ItnimeP.rogress, presented in Fig 2b. In generated dotted chart [XXX], each row corresponds to th th The color of the dot is associated with activity type. One may notice an inexactly one process instance and each dot corresponds to activity. Process instances creasing number of requests reported over time. The wide variety of the process are sorted from the top according to start time. The color of the dot is associated with instances duration, another typical characteristic of weakly structured spaghettiactivity type. One may notice an increasing number of requests reported over time. It tsyigpnicmalencht"ar(arcetperriesstiecntoefdspbaygbhleuttei-dlioktes)paroncdes\sCeos.mInptleerteesdti/nCgl oissetdhe" (farecptrthesaetnetxeedcugtoioldn odfotsQ)uaecuteidv/iAtiwesaiistinsgyncAhsrsoignnizmedenatcr(obslsuemadnoyt) paroncdesCs oinmsptalentceeds/C.losed (gold dot) activities execution is synchronized across many process instances.
a)
b)
According to the event log, Volvo IT is divided into 24 organizations, 594
6S4T,3s,8%an),dA6205(1f2u5n0c8t,io1n9,0d9iv%is)i,onBs.(4T6h2r3e,e7,o0r5g%an).izFartoiomns59a8reSeTssp,etchieallfyouarcotifveth: eCm e(4x1ec.1u8te9 inacttoitvailtiaelsm,o6s4t.3380%%)o,fAw2or(k1:2G.59078,(o1r9g.a0n9i%za)t,ioannCd, B74(646.6a2c3ti,vi7t.i0es5,%1)1.,3F9o%u)r, GST96s (eCx,e5c9u9te9,a9l,m15o%st),3S04%2 (oCf ,w4o3r8k2:, 6G,6997%()7,.4G62630ac(Bti,v1it6ie6s1,, 21,15.33%9%).)T,hGe9t6hre(5e.m99o9s,t a9c.1ti5v%e)f,unSc4t2ion(4.d3i8v2is,io6n.6s9i%nc),luadne:d VG32_320 ((310.965601,, 24.75,23%3%).),TAh2e_t1hr(e9e97m7,os1t5,a2c2t%iv)e, E_10 (4527, 9,41%).
st (70,26%) activites is performed in the 1 support line. In particular this takes place in
SL are Sweden, nd rd
Poland, France, India. Finally, “2 line – the most active countries include Sweden, nd rd
3 ” support line operation is handled by France – 39 activities (0,06%). function divisions are V3 2 (30.950, 47.23%), A2 1 (9.977, 15.22%), and E 10 (4.527, 9.41%).
The geographical distribution of activities is presented in Fig. 3. 46.042 (70.26%) activities are performed in the 1st line, mainly in Sweden, Poland, Brazil, and India. The most active countries in the 2nd line are Sweden, India, Poland, and Brazil, performing 16.541 (25.24%) activities. 2.911 activities (4.44%) are performed in the 3rd line, mainly in Sweden, Poland, France, and India. Finally, \2nd 3rd" support line activities are handled in France, performing 39 activities (0.06%).
Fig 3.iFtGiiegs.ep3oe.rgfGorreamopgerhdaipicnhiaeclaacldhdiicssottrurinbitubrtyui.otnioofnacotifviaticest.iNvuitmiebesr.s Ninduicmatebtehresvoinludmiecoaftaectvivo-lume of activities performed in each country.
Incident Handling event log. Probem Handling process is much more structured in comparison w2.i2thThIensPtraonbcleemHHaannddlliinngg epvernotcelosgss. Still it cannot be called Lanagia-like process. In FiTgh.eZPZroZblaemthHeanfdullinlgbperhocaevssi oisrmruecchomrdoreedstirnuctthureedcilnocsoemdpearviseonntwliothgthise presented in a form of theIncsotanntcreoHla-fnldoliwng pdrioacgesrsa,malt.hoIutgihsitesatisllycatnonontobteiccaelletdhlaatsatghnea-lnikue mprboceerss.of performed
In Fig. 4a), the full behavior recorded in the closed event log is presented in a activities andformtheof tnhaeucomntbroelr- oowf dipagorsasmi.bAlelthtoruagnhstihteionnumsbaerreof pseirgfonrimfiecdaancttilvyitiessmaller when comparing toatnhdethIenncuimdebenrtoHfpaonssdiblilentgrapnrsioticoensssa.reHsiognwi eceanvtelyr,smdaoltletre din cthheaPrtropbrleems ented in Fig. ZZZb confirmHsanrdeling process uthnasntrinu cthtuerInecdidcenhtaHraancdtelirng processp,rthoecdeostst.ed chart premaining of the sented in Fig. 4b) illustrates the still unstructured character of the Problem a) Handling process. b)
A comparison of the Opened and Closed Problem Handling processes is presented in Tab. 1. The typical behavior for both logs is presented in Fig. 5.
In Fig. 5, the numbers associated with activities or transitions indicate the number of process instances passing through a particular activity or transition. The comparison of the Opened and Closed Problem Handling process is presented st hours) Most actiIvnecsiudpepnotrt Hlinaensdling2ned v(6e8n,t14l%og),. 3Prdro(2b8e,m92%H)a,n1dling proc2endss(6i4s,2m7%uc),h3mrd(o3r0e,5s2t%ru)ctured in comparison with I(n2s,9ta3n%c)e Accepted/In Progress (49,0S9t%ill) it canCnoomt pleted/Closed (L2a5,n9a7%gi)a-like
Activities CHomanpldetleidn/gClopsreodc(e16ss,4.6%) Accepbte%ed/)Ic,naGPllreogdr(e5s,s89(46%) Most actipvreoscuepspso.rItnteFamigs. ZZGZ4a2th(7e,3f6uQ%lule)bu,eeSdh3/Aa3vw(i7ao,i1tri9nrg%ecA)ossridgendmeinnt theGc1lo99sQe(ud9e,u8eev6de/Anwtaloit8ign8gisAspsrieg%nsme)nentetd in Most actiavefoorrgmanoizfatthioenscontrCol(-4f7lo,8w9 %(d1)6i,a,2Ag9r2%a(m)26. ,I0t3%is)e,aGs4y to noticCe (t5h0(a1,t528t,9h%5e%),n)Au2m(b2e4r,4o6f%p),eGrf3ormed activities and the (1n4a,u1m2%be)Arccoefptepdo/Wssaiibt(l9e,1%) C_6 A(s8icg,c7ne1pi%tfeid)c,/aAEns_tsl1iyg0ne(d8,(576,1%1%))
E_10 (14A,1c2c%ep)t,edA/A2_ss1ig(n1e2d,2t(r99a,%0n2s)%it)ions (a9r,e86A%c)cepted/Wait (7,s5m5%al)ler when Most funcctoiomnpdairviinsigonto the Incident HAsasnigdnleidn-gAspsrigonceeds(s0.,0H4%ow)eever, doCtotmedplectheda/rCtapnrceelsleedn(t0e,d06i%n)Fig.
Z6ZMZbinicmoanl fdbiurirmgantisoienrwemPaainszinkgiewu0nimcsiztlrliuasencctdounrWdesdilclyhaPraiccatredr of the pro1cmeinsust.e 22 seconds
Z
In Fig. MDaDximDa,l dnuuramtiobners assoc4iayteeadrs 3w00itdhayasctivities and tran3s5i6tidoanyss2i2nhdoiucrsate number of process ain)Mstoastnaccteivse spuapsposritnlignesthrou2ngdh(68a,14p%a),r3tirdc(u28la,9r2%a)c,1tistvity an2dnd (t6r4a,2n7s%it)i,o3nrd .(30T,5h2e%)general
b) conclusionMoisst atchtiavte saulptphoortutegahmsthe(G24s,92i3z(%e7,)3o6%f)t,hS3e3 l(7o,g19 %is) differentG,1t9h9e(9,c8h6%ar),aGc8te8r(i5s,8t9ic%)of the registered Mboeshtaacvtiivoerorigsansiizmatiiolnasr foCr b(4o7,t8h9%ev),eAn2t(2lo6,g03s%.), G4 C (50,58%), A2 (24,46%), G3
(14,12%) (9,86%) a) Most function division E_10 (14,12%), A2b_1) (12,29%) C_6 (8,71%), E_10 (8,56%)
process instances passing through a particular activity and transition. The general conclusion is that although the size of the log is different, the characteristic of the regiFstiFegr.iegZd.Zb4Ze.h.LaLveeivoverelliosofsfsismttrruiulcactrtuufroraarttbiiooonnthooeffvttehhneetCClollogossse.edd PPrroobblleemm HHaannddlliinnggpprroocceessss.
Tah)e comparison of the Opened and Closed Pbr)oblem Handling process is presented in Tab. WWW. The typical and the most frequent behavior for both event logs is presented in Fig. DDD.
Events 2351 4755 2 QuPeroscteisosinnsstances 819 1166
No. variants 183 236
The four aspectFsigh.av5.e TbyepeincailCnbovmeehpsaltevitegidoa/rCteloodfse(ddau)roipnegnethdepreovbelnemtCoslo,mg(pble)atencdal/oClsyloessdiesdp:ropbulsehmst.o front strategy, ping pong behavio(hr2o,1u,r4ws9)%ai,tm- ueasnedruration: 49 days, 1 hours)process conformity sub-status usa g(3e8,,81%, mean duration: 49 days, 6per organizSaTetciho2onendg.vQearnuiaenertastlicoonnsclusionAcicsepttheda/tIn, aPrlotghroesusgh the size of tAhcecelpotegds/IinsPdroigreesrse,nt, registered typical behaviors are(1s5i,0m2%ila,mreiann dburoattihon: O0)pened anAdccCepltoeds/eWdait, Completed/HClaonseddling Problem prTohceesfsoeusr. aspects have been investigated during the event log analysis: push to front 2.1 PushsttroatFegryo,nptiSngtraptoenggybehavior, wait-user sub-status usage, process conformity per organization.
In gener3al, QInuciedsetnitoHnasndling process follows the push to front strategy. In Fig 4, the diagram describing the most frequent variants of Incident Handling process is presentedT.h2N.e1otfPofuuurslhlbatoespheacvtinsotrShrtearvcaoeterdgbyeede nin tihnveeesvtiegnattelodg disucrianpgturtehde inevthenetdilaoggraman.aAlyssias:
Fro general rpuulseh, tthoe firnocnitdesntrtahteagnyd,lipnigngi-spponergfobrem-ed ipnasstheed t1ophsrat2pdIovnnricliaedieoggsnsareeresn,ann.mdetcweroNda3anl.dir,ufdtNeo-ImslnuroiccmnsbtrieiefedbirurteiilynnslsgotubpsfHbeemth-rahsapanetovlradlrioeltogmiucrrna.esognrsseisutpczsoraofeirtrn,dcieoeseqasntdunas.eindfnnocttlehlsovewaersviaetnhntetslpoougfshisItncocaipfdrteounnrettdsHtirnaanttehdgeliynd.giIangprFraoimgce.4sA,stshiaes general rule, the incident handling is performed 3.i1n Pthues1hst tloinFe.roNnutmSbetrraotfegpyrocess instances
passed to 2nd and 3rd line is smaller.
In general, the Incident Handling process follows the push to front strategy.
The most frequent variants of the Incident Handling process are presented in Fig. 6. As a general rule, incident handling is performed mostly in the 1st line. The number of process instances passed to 2nd and 3rd line is smaller by at a least an order of mag- Fig. 6. Typical incident handling. nitude. Statistic
To understand the exceptions to the push to front strategy, we have performed a two-step analysis. First, exceptions to the push to front strategy are studied from the product perspective, i.e., products that follow and do not follow the strategy are indicated. Second, Volvo IT organizational structures are checked with regard to their respect of the push to front strategy.
Products. All the products were analyzed in terms of involvement of support lines with two approaches: (1) analysis of transitions among support lines, (2) analysis of number of activities per lines.
In the rst approach, 12 methods of incident handling were distinguished depending on the set and ordering of involved support lines. Correct methods, i.e., methods supporting the push to front strategy, include: process instances fully performed in the 1st line (marked as "1"), process instances started in the structures are validated on varaioreu:s l2e,ve3l,s f1o-r3b,e2in-g1,in2l-i3n,e with the strategy. line (1-2-3). Incorrect variants 2-1-3,P3r-o1d,u3ct-s2.,A3l-l1t-h2e.products recorded in the Incident Handling eTvaebnlet l1o.gAwctievrietyacnoaulnytzfeodr
In iFnigte5rma,sporofdiuncvtoslvtehmatenatreofweslulppaloirgtnleidnesw. itPhropduuschts- were cporomdpuacrtsed424usainndg66t0wo to-froanptpsrtoraactheegsy:(a1r)eanparleysseisnotefdtr,ani.sei.t,iofnosratmhoonsge spurpopdourtcltisnes, (2) analysis of number of st the moacsttivoiftiepsropceerfsosrminesdtainncevsariisoufuslliyneesx. eIcnuthede ifnirsttheap1Fpirgo4a.cThy,pitchaelintcwideelnvtehamndelitnhgods of line. Tinhciidsesnet thainndclliundgewsepreroddisuticntgsu:is4h2ed4,de6p6e0n,di3n8g3o,n25th3e, set and ordering of involved 5866, 4su9p4Zp.obritghnleiinetewws.oPC,aomsrzrokecsiettwmfriceeztqhuaoendnsdtilWnyclieullxdyeec:Pupictreaodrcdepssroindsutcantsces fully performed in the 1st
T 7n(4f13uos26mrt44wb,aa2ll(eniia16r-rnndn1-d6ee2o-de03)f((fd,o)m,12par3-atw2a7rc-roo1-rt2ka3ice,e8vtre)3dh.dis-wtseI2aeianeo,cdsi2cnst3r“oisntt-v1pthr1odar”ie-ent)2rticaa,cfe.htenopssevcrrdomalsa2ocrtlranieeasaerseddtntseretpdsflriinoelnaitosrrnrtoefoaetotn:kh(thrc1.eh2mees-,esI21ee3ns)3dst,p,traa1rTdpnrio-ntad3redlob,diufntcl2ohiecee-nrte1wss(s1t,ah11i2rien-sdn-t2th31ese-,dselt3taian)acton.nehdcIthnesffuosecopror2swprtnprdaoaerrrraodcttntededldduimnlctaieotetnettir4hhstet2hoop4ed2rtnseh1dasesanleitr3nndereatd:en6d2d6,0– re3s,p1e-c3t,ivI2ne-l1Fy,i.g2H-53aa,,n2pdr-lo1ind-g3uc,it3sn-ct1hid,ate3na-r2tes, w3fo-e1rll-t2ahl.iegsneedsewt itohf ppursohd-uctsTianpbrcloeldu1u.dcAtesc4tia2v4litsyaoncdomu6n6te0ftohrods 1-2 and 2.toI-nfrcoanstestorafttehgeysearperpordeusecnttse,dt,hie.ei.n,cfoorrrtehcotsempertohdoudcsts2 is performed very rarely. a) tlhineem.Toshtisofseptroicnecslsudinesstapnrcoeds uisctfsu:ll4y2e4x,e6c6u0te,d3ib8n)3t,h2e513s,t 566, 494. The two, most frequently executed products (424, 660) are worth a closer look. In Table 1 the number of activities performed for these products in each support line is presented – 7364 and 2728 activities are performed in the 1st line for product 424 and 660 respectively. Handling incidents for these set of products include also methods 1-2 and 2. In case of these products, the incorrect methods 2 is performed very rarely.
In Fig 5b, products are ordered according to number of process instances following variaInnt F1-ig2.. 7P,ropdruocdtsucatrse thstailtl awreellwaellliganleidgnwedithwiptuhshp-utos-hf-rtoon-tfrostnrtatesgtryatbeugty oafrteen foprrwesaerndteeddt,oi nthcelusdeicnogndprloindeu. cTtshifsosrewtihnicclhudmeossptroofdtuhcetsp4r2o4c,es5s4 2in, s6t9a8n,c2e5s3a,r6e6f0uallnyd 2e6x4e.cuted in the 1sFtigli5n.eP,rosduucchtsawse4ll2a4li,gn6e6d0w,3ith83p,us2h5t3o,fr5o6n6t,starnatdeg4y9:4 (cf. Fig. 7a)).
PrTohduecttwso mfar)oespqtrod
ufernuetcqltyusefonltrloelyqwueinixrgein1cguvtaeridathnpet;rob)dpurcotdsu(ct4s2f4olalonwdin6g610-)2 avraeriawnotrth a closer sluopopko.rt7.o3f6t4hean3drd l2in.7e2a8reacptrievsietinetsedarine Fpiegrformed in the 1st line for products 424 6a.nAd gv6ra6erI0ianatnrF\teei1gsx1-pa-252emb".c,tppaPirlvnreooedddlouyu\fc.c2ttssHs"ua.acarenrhIendoarlsidctnpiealgrlrseeodwidnoaeucclfliccdtotaehrlidniesgitsnnsegedfptoorrwonuidtthmhuebcspteeusr,soshfte-httpoer-oofcfrionepsncstrooirsndrtsreutaacctntetcgsmeysienfbtocuhlltluoooddwfeitnenaglso methfoordwsarded to the second line. This set includes products 424, 542, 698, 253, 660 an\d2 is perfo2r6m4.ed very rarely.
In FPirgo.du7cbts), pfrreoqduuencttlsy arreeqourirdinegredthe accorsdupinpgort oofththee n3urdmlinbeeraroefpprersoecnetsesd iinn-Fig stanc6e.sAfgorlleoawteinxgamtphlee o\f1s-u2c"h mapertohdoudc.t is Products are still well aligned with push-to-front strategy but often forwarded to the second line. This set includes products 424, 542, 698, 253, 660, and 264.
Products frequently requiring the support of the 3rd line, e.g, product 698, are presented in Fig. 8. It is noticeable that for most products in this set, e.g., 604, 295, 617, 611, Fig. 8. Products following the \1-2-3" the incorrect-to-correct-executions ra- method. tio is important, with a signi cant presence of the \2", \3", and \1-3" methods.
In Fig 9, the set of 30 most frequently supported products (corresponding to 53% of all the activities) is presented (without products 424 and 660 discussed earlier). 698. It is noticeable that for other products in tPhirsodseut,ctis.e., 604, 295, 617, 611, the incorrect to correct executions ratio is higher. In particular the presence of 2, 3 and 13 is clearly Sviuspibpleo.rt Line 13 253 267 321 383 453 494 544 566 698
In Fig 7, the set of 13s0t m7o1st3f1re6q0u8en7tl6y9ap6p4e0ar8in3g9 p3ro2d5u1ct2s0i8s p3r1e9sen6t5e1d (8p7ro7ducts 424 and 660 have been removed f2ro6m1 t2h2e9figu9r9e as4t9hey are discus2s0e1d earl-ier). These
2nd 132 109 213 424 products correspond to3r5d3% o1f9activit-ies 3 - 2 3 - - - 193 recorded in the event log. Fig 6. Products following 1-2-3 variant execution. a) methods for each selected product. On the right side, the percentage of wrong executionsTfhoer meaacjohristyeleocftperdodpurcotdsufrcotm. Fig 7a are well aligned with push to front strategy.
This is confirmed also by Table 2. In Table 2 the activity count for the 10 most frequent products from Fig 7 is presented – the majority of activities are performed in theT1hstelimnea. jority of products presented in Fig. 9a) are well aligned with the push to front straTteagbyle. 2T. Chiosmcplaariismon iosf tshuepm- ost frequently executed products ported also by Tab. 2 in which the number of activities per support line and per product is presented for the 10 most supported products from Fig. 9. The majority of activities are
In Fig 7a, there are two significant exceptions that do not follow the preferred pesrtrfoartemgye:d 6i0n7t,h2e431.stThliinseo.bservation is confirmed by Fig 7b. In Fig 7b columns
indInicaFteigt.he9ap)e,rctewntoagseigonfiwcraonngt eexxecceupti-ons. The percentage is especially high for tiopnrosd|ucptsro6d0u7catnsd602743a. nAd 2hi4g3h|er dnoumnboetr of incorrect executions is also true for fopllroowdutchtse: 2p3u5s,h23t6o, 8fr0o2n.t strategy. This obseTrvhaeteioxanmipslecoonf thremwedronbgyinFciigd.en9tbh)a.ndling is presented in Fig 8. The diagram InprFesiegn.ts9bth)e, most hfreeiqguhetntobfebhaavrisorinfodri-product 607. All the activities during the caitnecsidetnhtehapndelrtihcnegenatraegpeerfoofrmwedroinngthee3xrde-line. cutions. The percentage is especially high for the formerly mentioned products 607 and 243 as well as products 235, 236, and 802. The most frequent behavior for product 607 is illustrated in Fig. 10: all the activities during the incident handling arFeigpe8r.foCrmomedmoFnigb. e1h0a.viCoormimnoinncbideheanvtiosrhianndinlciindgenftosr product 607 in the 3rd line. handling for product 607.
The products that follow incorrect methods of executon are presented in Fig 9 and Fig 10. In Fig 9 products are ordered according to the absolute number of the wrong executions (product 424 is skipped as it was discussed earlier). The set of produts presented in the two figures is quite similar. Products that do not follow the push to front strategy are: 607, 243, 54, 818, 82, 319. 10 Zbigniew Paszkiewicz and Willy Picard
Fig 8. Common behavior in incidents handling for product 607
The pPrroodduucctststhfaotllofowlilnogw iinnccoorrrreecctt emxeetchuotdiosnofmeextehcoudtsonaraereppreresesnenteteddiinn FFiiggs.9 1a1nd Fig 1aw0nr.doInn1g2Fe.ixgIenc9uFtpiigroo.nds1u1(c,wtspitrahorodeuuotcrtpdsreoaredrdeucaotcrdc4oe2rr4eddidniagsccutcoossrtedhdienegaabrtsloioelrtu)ht.eeTnahbuesmosbleuettsreoonffuptmhreobdewurrcootsnfg execpurtieosnenste(pdroinduthcte 4tw24o isgusrkeisppaered saimsiiltarw. aPsroddisuccutssstehdateadrolienro)t. fTolhloewsettheofpupsrhodtuots presefrnotnedt sintrathteegtywaorfeigpurroedsuicstsqu6i0te7,s2im4 3il,a5r.4P,r8o1d8,uc8t2s, tahnadt 3d1o9n.ot follow the push to front strategy are: 607, 243, 54, 818, 82, 319.
FFigig. 91.1P.rPordoudcutsctwsiwthitthhethheighhigehstesatbsaoblsuotluetneunmubmebreorfoifn cinocrorrercetcptrporcoecsesssinisntsatnacnece executions. executions
The typical behavior of products 818 and 54, exceptions to the push to front strategy, is presented in Fig. 13: all the activities are performed in the 2nd and the 3rd line. The numbers associated with activities and transitions indicate the numbers of process instances the activity and transitions were performed in.
Products most frequently pushed to the 2nd line and fully executed in this line are products 243, 818, 54, 591, 82 and 319 (Fig. 14a). Products frequently handled in the 3rd line are 607, 568, 604, 5, 80, and 609 (Fig. 14b).
131 products do not appear in the event log often but are systematically handled in an incorrect way, with 322 such incorrect executions. Among these 131 products, 9 are always handled only by the 3rd line (80, 70, 74, 102, 578, 89, 75, 57, and 610), and 71 products are handled only in the 2nd line (e.g., 82, 319, and 762). Fig 10. Products with the highest percentage of incorrect process instance
executions
As an example, the diagram presented in Fig 11 captures the common behaviour when handling incidents of products 818 and 54. All the activities visible in the diagram are performed in the 2nd and the 3rd line. Numbers associated with activities and transition indicate number of process instances the activity and transition was Fig.F1i2g.1P0r.oPdruocdtsuwctisthwtihthe hthigehheisgthpeesrtcepnetracgeenotafginecorfriencctoprrroeccetspsrioncsteasnscienesxtaencucteions. performed in. executions
As an example, the diagram presented in Fig 11 captures the common behaviour when handling incidents of products 818 and 54. All the activities visible in the diagram are performed in the 2nd and the 3rd line. Numbers associated with activities and transition indicate number of process instances the activity and transition was performed in.
FFigig. 1113..CToympmicoanl behhaavviioorriinni ninccididenentthahnadnldinlignfgorfoprropdruocdtusc8t1s88a1n8da5n4d 54.
Products most frequently pushed to the 2nd line and fully executed in this line are: rd V2o43lv,o818I,T54,o5r9g1a,n8i2zaantdio3n19al(Fsigtr1u2ca)t.uPrreosd.uctAs frequenhtylphaontdhleesdisinitsheth3atlifnuenacrteions rst a6r0e7a,t5t6a8c,h6e0d4,t5o, 8li0neasn.dI6f0t9h. is hypothesis were correct, a given function would be alwaa)ys performed on a given line. Then, lbi)nes would be considered as separated by functions.
However, the analysis of the incident dataset leads to the disproval of this hypFotighe1s1is.:Cfuonmcmtioonnsbaerheamviaoirnilny isnccaitdteernetdhaamndolningglifnoersp,raosdpurcetssen8t1e8daonndt5h4e chord diagram in Fig. 15.
Products most frequently pushed to the 2nd line and fully executed in this line are: 243, 818, 54, 591, 82 and 319 (Fig 12a). Products frequentl handled in the 3rd line are 607, 568, 604, 5, 80 and 609.
Products most frequently pushed to the 2nd line and fully executed in this line are: 243, 818, 54, 591, 82 and 319 (Fig 12a). Products frequentl handled in the 3rd line are 60712,568, 6Z0b4ig,n5i,e8w0Paansdzk6i0ew9.icz and Willy Picard
In a general manner, chord diagrams shows relationships among groups of entities. Each block on the exterior circle represents an entity. The links between entities represent shared characteristics of the entities. Additionally, the size of the link area is proportional to the number of common characteristics of the entities. The ticks on the exterior side of the circle aim at providing guides to the number of common characteristics.
On this diagram, four lines \1st line", \2nd line", \3rd line", and \2nd-3rd", matching the line names provided in the provided data sets, are represented by 4 separated blocks on the outside circle. Two blocks are linked according to the number of functions that they both perform. A particular case is the case of functions performed by a unique line. In this case, an area attached to a single block is drawn.
From the chord diagram, one may identi ed that no function performed by the 1st line is performed only by the 1st line. 13 functions performed by the 1Fsrotmlintheeacrheorpderdfioargmraemd, obnye tmheay 2nd libidyneenthtietfio1eosdt,litnh1ea0tisnfupoenrfcfuotnirocmtnieosdnoapnreleyrfobpryemrthe-de forme1dst lbinye.t1h3e functions perforfmunedctbioynthse
3rd line. 11 are p1esrt floirnmeeadre bpeortfhormbyed tbhye t2hned2nadnlidne 3rd lintoeos,.1T0hfeun\c2tinodn-s3arrde"pleirnfoerpmeerdfobrymthse 1 fun3crtdiolinnei.n11cofmunmctoionnswaitreh ptehrefor1msetd line, baontohtbhyerthefu2nctainodn3rwdilitnhes.thTehe2“n2dnd
nd rd line, a3nd“a llianset pfuernfoctrmioend w1ithfutnhcteio3nrdin loinnley. bOffcuuoynnnmccltmytthiioooe8nnn2fwnwuwidnitithtchhltitnihtotehehn,ees2an3n1adrsdrdltein3lleipin,nfeeuera.,nfnocdOartnmniaoolyetnlhdases8rt are pefurnfocrtimonesd aorenlypebrfyortmheed3ordnlylinbey. Athe summ2andrylinoef, atnhde3nfuunmcbtioenrsoafrefupnercftoiromnesd perforomnlyedbybythetw3ord lliinnee.s is presented in FigF.i1g51.3F.uFunnccttiioonn sshhaarirnigngacarocsrsolsinselsi.nes.
functions performed by two lines is presented in Table 3.
An hypothesis is that organization lines are spread amongst lines, assuming that the competences are scattered within each organization line among all the lines. If this hypothesis were correct, each organization line would be operating on each line and no organization line would be confined to a single line.
The analysis of the incident dataset leads to a partial confirmation of this From the chord diagram, one may identified that no function performed by the 1st line is performed only by the 1st line. 13 functions performed by the 1st line are performed by the 2nd line too, 10 functions are performed by the 3rd line. 11 functions are performed both by the 2nd andT3arbdlleine3s.. FTuhnec“t2iondn-s co-performed by lines 3rd “ line performed 1 function in common with the 1st lin1e,st alninoeth2enr d line 3rd line 2nd-3rd line function with the 21nsdt lliinnee, and a0last function with the2n3drdli nliene. On1l3y 8 functions are perfo3rrmdeldineonly b1y0the o2nndlylinbey, tahned33rd2fluninndec-.t3irodnslianree perform1 ed 13 10 1 8 11 1 11 3 1 1 Fig 131. Function shari0ng across lines.
A summary of the number of functions performed by two lines is presented in Table 3.
Another hypothesis is that organization lines are spread amongst lines, asTable 3. Functions co-performed by lines suming that the competences are scattered within each organization line among all the lines. If this1shtylinpeothesis we2rned clionerrect, each3rdolrigneanization 2linnde-3rwd oliuneld be oper1asttilninge on each l0ine and no orga13nization line w1o0uld be con n1ed to a single line.2nd line 13 8 11 1
T3rhdeliannealysis of t1h0e incident data1s1et leads to a p3artial con rmat1ion of this hypoth2ensdi-s3:rdorlignaenizatio1n lines are mai1nly participatin1g in many lines0, as presented on the chord diagram in Fig. 16. On this diagram, four lines \1st line", \2nd
An hypothesis is that organization lines are spread amongst lines, assuming that line"t,he\c3ormdpleinteen"c,esanarde \sc2anttde-r3edrdw"i,thminaetacchhinogrgtahniezalitinoen nlianmeaems opnrgovalildtehde liinnetsh.Ief pthriosvidehdydpoatthaessiestsw,earerecroerrpercet,seenactehdobrgyan4izsaetpioanralitneedwboluolcdksbeonoptehraetionugtosindeeaccihrclline.e Tanwdo blocknos oarrgeanliinzakteiodnalcinceorwdoiunlgd btoe ctohnefinnuedmtboearsoinfgoler glianne.ization lines participating in these lTinhees.aAnalpysairsticoufl athrec ainsceidiesntthdeactaasseet olefaodrsgatonizaatpiaorntiallinceosnfpiremrfaotriomninogf othnisa singlheyplionteh.esIins: tohrigsanciazsaeti,oannlinaerseaaraetmtaacinhleydptaortiacipsaintignlgetoblmocaknyislindersa,wans.presented on Fthroemchtohrde dchiaogrrdamdiiangrFaigm1,4o.nOmnatyhis
st nd idendtiifaygrtahma,tfo9urolringeasn“iz1atliionne”,li“n2es lainree”, activ“e3rdo nlinbe”o,thandth“e2n1d-s3trd“a,ndmatthchein2gntdhe linesl,in5e onragmaensizpartoivoindeldiniens tahree parcotviivdeed
data sets, are represented by 4 separated on bboltohcktsheo1nst tahned otuhtesid3erd cliinrcelse,. aTnwdo 7 orgbaloncikzsation lines are active on btooth
are linked according the the 2nnumdbaenrd thoef 3rdorlginanesiz.aOtionne orglian-es nizatpiaornticliipnaetinagctinivteheosne ltihnees.\A2npda-r3tircdu"lar line cisasaelsios tahcetivcaeseonofthoerg1asntizlaitnioen, alinn-es othepreorfnoermoingtohne a2snindgleinlein,ea.nIndthaislacasste, one aonn atrheae atthtaircdhedlinteo. aThsienrgelfeorbel,ocokr-is ganidzraatwionn. lines tend to participate in more tFhraonmo ntheelicnheo.rd diagram, on may
However, 6 orgaorngiaznaitziaotnionlinliensesopa-re identify that 9 erate only on the 1st line and 6 orga- FigF.ig1164.. OOrgaannizizaatitoinonlinleisnaecsroascsrolisnseslines. nization lines operate only on the 2nd lines. Therefore, the hypothesis of a systematic spreading of organization lines across lines is not fully acceptable.
A summary of the number of organization lines operating on various lines is presented in Tab. 4.
Another hypothesis is that process owners are attached to a given line. If this hypothesis were correct, a given process owner would performed only on the 1st active on both the 1st and the 2nd lines, 5 organization lines are active on both the 1st and the 3rd lines, and 7 organization lines are active on both the 2nd and the 3rd lines. One organization line active on the “2nd-3rd” line is also active on the 1st line, another one on the 2nd line, and a last one on the third line. Therefore, organization lines tend 14 to parZtibciigpnatieewinPmaoszrekitehwaniczonaenldinWe. illy Picard
However, 6 organization lines operate only on the 1st line and 6 organization lines operate only on tThaeb2lned lineOs.rgTahneirzeaftoioren, ltihneeshyshpaorthinegsisacorfosas slyinsteesmatic spreading of 4. organization lines across lines is not fully acceptable.
A summary of the numb1esrt olifneor2gnadnizlianteio3nrdlinliense o2pnedr-a3tridnglinoen various lines is presented in Table 4.1st line lines.An hypothesis is that persons are
aTttahcehedanatolysais goivfenthleine.incIifdetnhtis dathayspeotthleesaidsswteoretchoerrdecist,parogviavlenofpethrsiosn hypwootuhldespise:rfaorlmaergdeonnluymonbetrheo1fsptlrioncee, s2snd owlnineersoris3rodpleirnaet,inagndonno mpearnsoyn liwnoeusl,d as oppreersaetnetoendtwono othremochreolridneds.iagram in Fig. 1T7h.e Oannaltyhsiiss odfiatghreamin c,idfoenutr dlaintaesset \1sletaldinseto",th\e2ndidsplrinovea"l, o\f3trhdislihnyep"o,thaensdis: \2nadl-a3rrgde"n,ummbaetrchoifngpetrshoensliinseonpearmateinsg proovnidmeadny lintehs,eas presented on the chord
in provided data sets, oarneld“it2inrhaneegdeps-r3ra“oerm1dus“stet,isnnliitdnFmeeeid”ag,tcbc1i“hry52ci.nnl4dOegl.snineTttephhw”aie,sro“ad3btliriaedlngodlercianbkmelsn”o,a,cafmakornesuedsr linkpreodvaidcecdoridninthgeto the numdbatear soeftsp,roar-e
provided cesrseporwesneenrtesdobpyer4asteipnagraotend tbhloecseksloinnetsh.e A particular case is the case of process oawreaoonacpuectearsorstiardtdtoaeiinpcnghcgeierroadctntoletino.ttghhaTeeowssenionnluagibmnlleseobisbcne.krlgosAlceokaflprieianspretelti:irhcnsuaekolnnenadsr lFiniges.F.i1g71.5.PPreorscoenssinvoowlvneemrenintvinollivneemsent in dracwasne.is the case of persons operating on a single line. In this case, an area attached to a siAngbloeubtlo7c0k0isprdoracwesns. owners are operating on both the 1st and 2nd lines. About 250 pArobcoeusts7o0w0 npeerrssoanrseaorepeorpaetriantgingonatbbootthh tthhee 11ssttanadnd2ntdhlein3ers.dAlibnoeust. 2A5l0mpoesrtso3n0s0 proacresospeorwatninegrsatabroethoptheera1tstinangdothneb3ordtlhineths.eA2l mndosta3n0d0 tpheerso3nrsdalrienoeps.erTathineg patrobcoethss owtnheers2ndopanerdatthineg3rdo nlintehs.e T\h2enpde-r3srodn"s olipneeraatirneg opnetrhaeti“n2gnd-o3nrd”thlinee1asrte loinpera(t5ingpraotcestsheo w1nelirnse),(52npderslionnes)(,22npdrloincees(s2opwenrseornss)),, aanndd tthhee33rdr dlinlein(e5 (p5erpsoroncs)e.ssThoewrenfeorrse),.
st Thmeroesftopreer,smonossatreporopceerastsinogwant etwrsoalrineeospoerrmaotirne.g on two lines or more.
HHowoweveevre,r,stsitlilllaalalarrggee nnuummbbeerr ooff pperrosocensssaorewonpeerrsatairneg oopnearastininggleolninae:saibnogulet 3li0n0e: li3n0e0,apnrdowlceoesrssksitnhogawnon1ne0lry0sopanerretshoewnso1sratkrleiinnweg,oorakbniolnyugtoo2nn5l0ythpoeenr1sthosents3liarndreeli,nwaeo.brTokhuinetgre2of5on0rley,poarnoccothersees aowbop2nuenedrtrssonasrearweorking only on the 2nd line, and less than 100 process owners are working only on the 3rd line. Therefore, a core of process owners are focusing only on a single line while a large number of process owners are operating in a cross-line mode.
A summary of the number of process owners operating on various lines is presented in Tab. 5.
Another hypothesis is that STs are attached to a given line. If this hypothesis were correct, a given ST would operate only on a given line, and no ST would operate on two lines. 1st line 1st line 22nndd lliinnee 3rd line 23nrdd-l3inrde line 1st line 2nd line 3rd line 2nd-3rd line cluding line a2)nmdo-3strdfr.equent behavior, b) process instances including line 2nd 3
In process instances where line “2nd 3rd” appears, only the 4th and the 5th lines appear, i.e., no 2nd and 3rd line are present (Fig 17b); line “2nd 3rd” interacts only with forwthaerd4sth tlihnee,wi.oer.,kretcoeitvhees 4antdh floinrwea(rcdfs. tFheigw.1or9ka)toalninde b4)th. (BFoigth17tah, eb)4;tbhoathntdheth4eth 5th lineasnidnttheera5cthtslinmesaiinntleyrawctisthmtahinely1wstitahntdhet1hset a2nnddthlein2en.d Alinsea. Acosnasceoqnuseenquceen,cSeTSsTfsrom thef4rotmh athned45ththanldin5ethblliunre tbhluer bthoerdbeorrdbeertbweetweneetnhtehe1s1tstaanndd tthee 2nnd.d.ThTishesrtreofnogrley, the linesuwmitmhartyheofptuheshnutmobferroonftSsTtsraotpeegrya.ting on various lines is presented in Table 6. exisctoenntcriebuotfesthteo 4dtihffiacnuldtie5sthwliitnhebsesintrgoningl ylinceonwtirtihbuthtees ptuoshdito cfuroltnitesstwraittehgyb.eiAng in
TAanboleth6e. rSThsybpyotlihneessis is that process owners are performing a given function, which is aligned with st line
a function-orienndted managemrednt. If this hypndothrdesis were 1 2 line 3 line 2 -3 line correct,st a given process owner would focus on a given function and specialization 1 line 201 34 0 0 would2bnde the rule for employees.
line 34 255 16 1 Therdanalysis of the incident dataset leads to a disapproval of this hypothesis.
3 line 0 16 91 0 Indeed2,nda vrdery small number of process owners are performing a unique function, -3 line 0 1 0 0 as presented on the chord diagram in Fig. 20. On this diagram, each block on the outside circle represents a function. Two blocks are linked according to the
An hypothesis is that persons are performing a given function, which is aligned numwbitehraoffupncrtoiocne-sosrioewntnederms apneargfeomrmenitn.Igf tthhies hayspsootchieastiesdwefurenccotriorencst,. a given person wFooruldalfmocousstoneaacghivpenaifrunocftiofunnanctdiospnesc,iailtizaetxioisntwsoaulpdrboectehsesruolwenfoerre mthpalotyepeesr.forms both fTuhnectainoanlyss.isThofe trhaereinecxidceenpt tdioantasiest fluenadcstioton aCdi5s.apPprovcaelssofowthniserhsyppoetrhfeosrism.ing funcIntdioened,Ca5vaerrye semitahlelrnupmerbfeorrmofinpgerDson2s, aEre8p,eorfrorEmi1n0g (acfu.nFiqiuge. 2fu0n)c.tiTonh, earsefore funcptreiosenntCed 5onistheeitchhoerrdrdairaeg,raomr irneFqiugi.r1e8s. aOnvtehriys dniaagrrraomw, eoafch block on thAe nouottshideer inexpertise.
circle represents a function. Two blocks are linked according to the number of teresting fact is that process owners performing function C 3 are performing persons performing the associated functions. most existing functions but C 5, D 2, E 8. Therefore, process owners performing
For almost each pair of functions, it exists a person that performs both functions. function C 3 seem to \avoid" functions performed by process owners performing C 5.
Fig 18. Functions performed together by employees. On the left side, the set of functions performed by employees performing C_5 are highlighted. On the right side, with a function-oriented management. If this hypothesis were correct, a given person would focus on a given function and specialization would be the rule for employees.
The analysis of the incident dataset leads to a disapproval of this hypothesis. Indeed, a very small number of persons are performing a unique function, as presented on the chord diagram in Fig. 18. On this diagram, each block on the outside circle represents a function. Two blocks are linked according to the number of persons performing the associated functions. BPI Challenge 2013 17
For almost each pair of functions, it exists a person that performs both functions. Fig 18. Functions performed together by employees. On the left side, the set of fFunigc.tio2n0s. pFeurfnocrtmioends bpyerefomrpmloedyeteosgpetehrfeorr mbyinegmCp_lo5yeaeres.hOignhltighheteledf.tOsnideth,ethrieghstetsidofe, tfhuenscetitoonfs fpuenrcftoiromnsedpebryfoermmpeldoybeyese mpeprlfooyrmeeisngpeCrfo5rmarienghiEgh_l1ig0hatered.hOignhltighhetreidght side, the set of functions performed by employees performing E 10 are highlighted.
Another hypothesis is that a given organization line is responsible for a given function. If this hypothesis were correct, functions would not be performed by various organization lines.
The analysis of the incident dataset leads to a partial con rmation of this hypothesis. Indeed, some organization lines tend to share functions with a large number of organization lines, while other organizations do either share their functions with a very limited number of organization lines, or perform functions in an exclusive manner, as illustrated in Fig. 21.
The rst group of organization lines sharing their functions with a large number of STs corresponds to the organization lines B, C, E, F, G1, G4, H, and V* (except V8). These organizations lines share their functions among themselves, which lead to the supposition that this organization lines are either performing a large range of generic operations or that they are very scattered organization lines having a very similar pro le. The presence of the large C organization line in this group is to be noted.
The second group of organization lines contains the organization lines A2, D, G2, G3, and V8. These organization lines don't share their functions with a large number of organization lines, but rather either share them with organization lines within this second group. The case of A, D, and V8 is special as these organization lines have speci c functions that are performed exclusively by these organization lines.
The case of the organization line C is interesting as on the one hand organization line C share functions with many organization lines, on the second hand, a large set of functions performed by C is only performed by C. manner.
The first group of organization lines sharing their functions with a large number of STs corresponds to the organization lines B, C, E, F, G1, G4, H, and V* (except V8). These organizations lines share their functions among themselves, which lead to the supposition that this organization lines are either performing a large range of generic operations or that they are very scattered organization lines having a very similar pro1f8ile. ThZebpigrensieewncPeaosfz kthieewliacrzgaenCd oWr gilalyniPzaictaiordn line is to be noted.
Fig 19. Functions in organization lines. On the left side, the case of the organization lineFiCg. i2s1.hiFguhnlicgthiotnesd.inOonrgtahneizartigiohnt lsinidese., Othnethceasleeftofsidteh,ethoergcaansiezaotfiotnhe loinrgeanAiz2a-is higthiolinghlitneed.C is highlighted. On the right side, the case of the organization line A2 is highlighted.
The ping-pong behavior refers to the situation when the incident/problem handling is performed by bouncing tasks between STs, organizations or functions instead of actually handling the request. In the simplest situation, the ping-pong behavior consists in repeated interactions among two parties. In more advanced scenarios, the ping-pong might be performed among larger number of parties in cycles.
In order to detect cycles, a social network analysis approach was taken. A graph, referred to as the cycle graph, has been built to capture cycles in process instances. The relation among parties exists if in one process instance parties perform activities after one another. When relations create cycles, they are counted. Only direct sequences of activities are taken into account. It is possible that one cycle appears in one process instance many times.
Three perspectives were taken for our analysis of the ping-pong behavior: (1) product perspective (tackling the question which products generate cycles), (2) cycle perspective (tackling the question which cycles are shared among products), and (3) organizational structure perspective (tackling the question which organizational units contribute to cycles).
Product perspective. In the Incident Handling process, the products with the highest number of cycles are products 424 (107 cycles), 38 (49 cycles), 542 (55 cycles), 802 (cycles), and 776 (27 cycles). On the other hand, some products are supported with a very limited number of cycles, but with one dominant cycle, e.g., product 258 (9 cycles, cycle G179, D8, G179 appears 26 times, cf. Fig. 22), product 295 (4 cycles, cycle G358, S9, G358 appears 6 times), and product 350 (7 cycles, cycle T17, D1, T17 appears 15 times). Fig 20. Frequent cycle G179, D8, G179 for product 258
Fig 20. Frequent cycle G179, D8, G179 for product 258
Cycle perspective. The most popular cycle is D4, D5, D4 and it appears 57 times in total foCrypcrleodpuecrtss:p2ec3t6iv,e6.97T,h7e9m1,os7t89po,p2u3la5r, 3cy2c8le, 5is42D,47,7D65,,1D548,a3n1d8i,t9a2p,p7ea3r6s, 312, 727, 799,5C73y2ctl9iem,pe3es2rsf7op.reTcpthirvoeed.duTicahtgesrm2a3om6st,ipn6o9pF7u,ilga7r.9c21y1,cl7ce8ai9ps,tDu24r3,e5sD,5t3h,2eD8,4p5rao4n2cd,eist7sa7p6ipn,es1atr5as8n5,c7e3s1ti8mw,eh9se2r,e this 736, 312, 727, 799, 329, and 327. The process instances in which this cycle has cycle iwn atostalrefocroprdroeddu.ctsC: y23c6le, 69b7e,tw79e1e, n789S,T235D, 4328a,n5d42,D7576,i1s58c,l3e1a8rl,y92,v7is3i6b,l3e1.2,Number occurred are presented in Fig. 23. The cycle between D4 and D5 is dominant. associa7t2e7d, 7w9i9t,h3a2c9t,i3v2a7te.sThaenddiatgraranmsitinioFnisg.in21diccaaptteurnesumthebeprroocfespsrioncsteasnsceisnswthaenrceetsh.is
The numbers associated with activations and transitions indicate the number of cycle was recorded. Cycle between ST D4 and D5 is clearly visible. Number process instances. associated with activates and transitions indicate number of process instances.
Fig 21. Cycle D4, D5, D4
Fig 21. Cycle D4, D5, D4
Fig. 23. Process instances containing the cycle D4, D5, D4.
products t6im97e,s,7p9r1o,d3uc1t8s,679879,)7.91, 318, and 789).
623, 790 (Fig. 22). The third of the most common cycles is D4, N26, D4 (50 times, Similarly, cycles G179, D8, G179 has been recorded 40 times for products 258, producStsim6 9il7a,r7ly9,1t,h3e1c8y,c7l8e9G).179, D8, G179 has been recorded 40 times for products 623, 790 (Fig. 22). The third of the most common cycles is D4, N26, D4 (50 times, 258, 623, and 790 (Fig. 24). The third most common cycle is D4, N26, D4 (50
Fig. 24. Process instances containing the cycle G179, D8, G179.
In Problem
Handling process, the ping pong behavior and cycles are not so common. Some cycles cIann tbheenPotriocbeldemin CHlaonseddliPnrgobplreomcessesv,etnhtelopgi.ng-pong behavior and cycles are not ThesoOcpoemn mPoronb.leSmo mseevceynctlelsogcains baectnuaoltliycefdreien oCflosed Problems event log. The Open
Fig 22. Cycle G179, D8, G179 cycPlerso.bTlehmiss ecvoennfitr mlosg ithseacotubsaellryvaftrioene otfhcaytclthese. This con rms the observation that structured.
There are btewhaovicoyrcalensd icnyctlhese Carelosneodt sPorocobmlemmons. eSvoemnte ProtbhleemProHbalnedmlinHganpdrolicnegssproiscesdsefiisnidteelynitmeloyremore structured.
In Problem Handling process, the ping pong There are trwstoccycycyclceleslecGsan1inb9e9tn,hoGetic2eC1dl,oinGseC1dl9o9sPerdaopPbrploeebmalersmss2e5vetnitmloegs.
log. The event log: G199, GTh2e1,OGpe1n99P.roTbhleemsse ecvyecnltesloagpipseaacrtsually free of
for products 97, 98, 96. Process instances a ected 25 times for producycctsle9s.7,T9h8is, 9c6on.fPirrmoscetshseinosbtsaenrcvaetsion that the
by this cycle includes only activities G199, G21 (cf. affected by this cyPcrloebilnemcludHeasndolninlgy apcrtoivceitsisesiGs1d9e9fi,nitely more
Fig. 25). Thestrsuecctuornedd. cycle G88, G92, G88 appears G21 (cf. Fig. 23). The second cycle G88, G92, G88
5 times for prodTuhcertes a3r6e3t,w7o9c3y,cl1es97in, 3th4e8.Closed Problems appears 5 times for products 363, 793, 197, 348.
event log: G199, G21, G199. These cycles appears among owwenreers and STs analyzed.
STs anaalyppzeeadrsw.5etriemes for products 363, 793, 197, 348.
Among owners, the case of Siebel is
Among process owners, the case of Siebel is partic
25 times for products 97, 98, 96. Process instances
Ppaerrtsypepcetrisvpee,ctthiveei.nItnertahcistiostnepa,mthoenginpterroaccteisosnowners and
G21 (cf. Fig. 23). The second cycle G88i,nGC92lo, sGe8d8Problems event log.
affected by this cycle includes only activFitiiges2G31.9C9,ycle G199, G21, G199 particularly interesting. Seibel has the highest degree in the cyclFeiggr2a3p. hC,ywclheiGch19m9,eGan2s1, G199
ularly interestinPga.rStyeipbeerslpheactsivteh.eInhthigishsetsetp,dtheegrineteerianctitohne that among all the owners participating into a cycle in a processininFCstiloagsn.ecdeP,2rS5oi.belbemeClsyiecsvltehnetlGog1.99,
cycle graph, wamhoicnhg omwneearnssantdhSaTts awmeroe nangalayzlledt.he process one connected to theAhmigohnegst nouwmnebrse,r otfheunicqausee o wofnerSsi.ebel is owners participating into a cycle in a process instance, In the representpaatirotincuolafrltyhientceyrecslteinggr.aSpehibpelrehsaesnthteedhiignheFsitgd.e2g4re,eeianPcthrhoenbcolydecmelersgerepavrpehns,etwnlhtosigcah. means
Siebel is the one connected to the highest number of person. The size otfheaatcamhonnogdaellrtehperoewsennetrss pthaertidciepgartienegoinfttohaecpyecrlesoinn:a tphreocbeisgsginesrtathnceen,Soideeb,el is the unique
procneusomsnbeoewcronnneercsct.eodntnoetchteehdignheosdtensu.mTbehreoreffuonrieq,ueaowbnigergse. r node represents a the higher the of
In the visualIinztahteiorenproesfenthtaetiocnyocflethegrcyacplehgprarpehseprnetseendteidninFFigig.. 2246,,eeacahchnondeordeeprerseepnrtse-a person involved in cycles with a large number of persons, while a small node
person. The size of each node represents the degree of the person: the bigger the node, reprseesnetnstsaappreorcseosnsinowvonlvere.d Tinhceyscilzees owfitehaachsmnaoldlenurmepbreersoefnptsertshoensd. egree of the process
the higher the number of connected nodes. Therefore, a bigger node represents a Tohwenberri:gthhtneebsisggoefrathnoedneordeep,rtehseenhtsigthheerbtehteweneunmnebsesrooff acopnenrseocnteidn nthoedecsy.cTlehoerfefore,
person involved in cycles with a large number of persons, while a small node G21,
G199 in
Closed giveonf npordoecebsestwoewen eTervhse,rbywrithgwhiltoneensaosdosefmsaa nlloddntehoredepoervereseernaptlrsletnshueemnbtbestewraeoenpfnrseohscoserotsefssatoppwearntshoenrb eiintnwtvheoeelnvcyecdle ionf evecryyctlweso wnoitdhesaignsrmaapahg.llrBaneptuhwm.eeBbneentrewssoefiesnpdnreeofsicnseesdresapsor wethsneeneratrstsio.thoef tihme pnourmtabnerceofosfhoartensotdpeathwsitthhrough a
given node between every two nodes and the overall number of shortest path between regard Ttoheinbfroirgmhatntieosns oflfoawns.odTeherephrigehseenrttshtehebebtewteweenennenssesosfofaa npordoec,estsheowmnoerrein the
every two nodes in a graph. Betweenness represents the importance of a node with important it is in terms of transmission within the graph. In Fig. 24, the darker a node,
cycle of graphr.eBgaerdtwteoeninnfoersmsaitsiodneflonwesd. aTshethheigrhaetriothoef tbhetewneeunmnebsserofofashnoodret,estthepamtohrse the higher its betweenness. In the case of the cycle graph, persons with a high through a giv eimnpnoortdanet bit eistwineteernmesvoefrtyrantswmoisnsioondewsitahinndthtehgeraopvhe.rInalFlign.u2m4,btheerdoafrksehr oarntoedset,
the higher its betweenness. In the case of the cycle graph, persons with a high path between every two nodes in a graph. Betweenness represents the importance of a node with regard to information
ows. The higher the betweenness of a node, are containing Siebel. Additionally, the degree of Siebel is the highest, and therefore, the node representing Siebel is the biggest.
Both the betweenness and the degrees of the nodes are fading down when going away from Siebel. The nodes are becoming brighter and smaller.
BPI Challenge 2013 21 Fig. 26. Process owner cycle graph. The cycle graph centered on Siebel. The size of the nodes is proportional to their degree, the brightness is proportional to their Fig 24. The cbyectwleeengnreasspchentcrealnittye.red on Siebel. The size of the nodes is proportional to their degree, the brightness is proportional to their betweenness centrality.
the more important it is in terms of transmission within the graph. In Fig. 26, The similarthaendaalrykesrisa hnoadse,bteheenhigher its betweefnonress. In the casevoafrtiheetycycle graph, performed STs. The of colors and sizes among nodes pcroorcresessopwonnedrsinwgithtoa hSigThsbiestwbeeingngeessr atrheaptaratmicipoantgingoiwnna elarrsg.e Tnuhmebcereonftral node is ST G97. ThiscyScTles pinavrotlivcinigpaatleasrgeinnummbaenr yof cpyroccleesss owwnietrhs, mwhailneyprvocaersisoouwsneSrsTwsit.hMa any cycles low betweenness are participatoinrggiannailzoawteironnumCb,ertohfecy1clsets with relatively few include G97, wprhoceersseoGwn9e7rs.belongs to support line and function division V3_2. AFlisnoalloy,thnoedresSaTre mlaiadyoubteasifnodlloicwaitneFdiga.s26t.hTohseececnotrnaltnriobduetSiinebgelt.oNoadpepsearance of cycles: G96 (olorcgataendioznattihoensmaller sctircle are fduirneccttliyocnonVne3ct_ed2)t,o Siebel.(Noordgeasnloizcaatteidoonn C, 1st line, the second circle Car,e d1irecltilnyec,onnnedcted to a node on thGe92rst circle. Therefore, function E_5),noGde2s7lo1ca(toedrgoanntihzeasteicoonndCc,ir2cle arleintwe,o fhuonpsctaiwoany fEro_m10Si)e,beGl,1th7a9t i(sotrhgeaynization C, 1st line, funchtiaovenpaVrti3c_ip2at)e, inGa2c3yc0le w(oitrhgsaonmiezoanteiownho hBa,s b2eennd plairntieci,paftuinngcintioancycEle _10), G40 (organization wAith Siebstel, or they have participate in a Dcy7cle (woitrhgasonmizeoanteiobnetwAee2n,themst alninde, function
Sie2be,l. 1 line, funcsttion A2_2), 1 A2_1), D2 (orgTanheizdaatrikoesnt nCod,e 1is Sielbinele.,SiefbuenlcisttihoenmoAst2c_e1nt)r,al Dno5de (inortgheanneitzwaotrikoonf C1st line, cycles in term of betweenness. Most of the cycles linking two process owners are containing Siebel. Additionally, the degree of Siebel is the highest, and therefore, the node representing Siebel is the biggest.
Both the betweenness and the degrees of the nodes are fading down when going away from Siebel. The nodes are becoming brighter and smaller.
A similar analysis has been performed for STs and is presented in Fig. 27.
The variety of colors and sizes among nodes corresponding to STs is wider that among process owners. The central node is ST G97. G97 belongs to organization C, the 1st support line and function division V3 2. It participates in many cycles with many various STs. Many cycles include G97.
Other STs are contributing to the existence of cycles (and therefore, to the ping-pong behavior): G96 (organization C, 1st line, function V3 2, in short C/1st/V3 2), G92 (C/1st/E 5), G271 (C/2nd/E 10), G179 (C/1st/V3 2), G230 (B/2nd/E 10), G40 (A2/1st/A2 2), D7 (A2/1st/A2 1), D2 (C/1st/A2 1), D5 f(uCn/ct1iostn/A2_1), D8 ((oArg2a/n1iszta/tiAon2 A1)2,, D1s4t l(inAe2, /f1usntc/tiAon2 A1)2,_G1)5, 1D(G44(o/r2gnadn/izuantikonoAw2n,). 1st line, function A2_1), G51(organization G4, 2nd line, function unknown).
FFiigg.252.7T. hSeTcyccylcelegrgarpahphc.enTthereedcyocnle Gg9ra7p.hThcenstiezreedofontheG9n7o.deTshies spirzoepofrtitohnealnotodes tihseiprrdoepgorreteio,nthael btorigthhtenierssdeisgrpereo,ptohrteiobnrailghtotnthesesir ibsetpwroepenonrteisosncaelnttoraltihtye.ir betweenness centrality.
This analysis can be concluded that the main organizations involved in cycles are
incidOenrgtaenveiznattlioogn.sOitnhveorlcvyecdlei-nprmonoesotrcgyancilezastiaornes CareaBnd,VA121., VSp7enc,ie.gc.a,lclyy,clteheC,cByc,le Ca m(2o9ntgimtehso)s,eC,oVrg1a1n,iCza(t6io4ntimapeps)e,aBr,eAd24,0B8 (t4i4mteimseins).the incident event log. CycleprIonneproorbgleamnizeavteinotnlsogarceycBle,sVa1re1,gVen7enra,tee.dg.b,ycyocrgleanCiz,aBtio,nCC(,2A92t,imG3esa)n,dCB,,Ve1.g1.,, C c(y6c4letiAm2e,sG),3,BA, 2A(22,6Btim(4e4s),tiCm, eAs2)., C (10 times), C, B, C (9 times).
In the Problem Handling event log, cycles are generated by organization C, 2A.32, WG3aiatnUdseBr, e.g., cycles A2, G3, A2 (26 times), C, A2, C (10 times), C, B, C (9 times).
In order to determine the use of Wait-user sub-status in Volvo IT, the Incident
Problem Handling event logs at all. In Incident Handling process, Wait-user substatus is always used with Accepted status. The activity Accepted/Wait-user was executed 4217 times. This accounts for 6,43% of all the activities registered in the event log. The Accepted/Wait-user activity appeared in 2495 (33%) process instances.
In order to determine in detail the patterns in usage of the Wait-user sub-status, the Accepted/Wait-user activities were analyzed with reference to: products, ST, countries, function divisions, line numbers, organizations and action owners
The \Wait-user" sub-status does not appear in Problem Handling event logs at all. Therefore our analysis focuses on the Incident Handling process.
The \Wait-user" sub-status is always used with the \Accepted" status. The activity \Accepted/Wait-user" has been executed 4.217 times, which accounts for 6.43% of all the activities registered in the event log. The \Accepted/Waituser" activity appeared in 2.495 (33%) process instances.
We have analyzed the \Accepted/Wait-user" activities with regard to products, STs, countries, function divisions, support lines, process owners, and organizationsP.roducts. The highest number of Wait-user sub-statuses was recorded for products the most frequently referred by incidents, i.e., 424, 660. In Fig. 26, twenty products Produwcitths.thePbriogdguescttsnu4m2b4er, o6f60subare asstsaotucsiatuesdageswiatrhe ptrheesenhteidg.h-The est nubmigbgeerr thoef c\ycWlea,itth-euhsiegrh"ersnuubm-ber sutcattsusweseouivsftb.ahl-sIusuntatabthte-Fusestia,egtlv.utahes2rneg8tuuer,aassttlatiwgooene.WvnuetamIrynuitbs-peueosrrreodordefors-futhbtoe- PRO2D7,8002 PR4O1D74,024 \Wait-sutasteurs" ususabg-estatcuousnta/nruembperre- of sentedp.rTodhuectslairngsetarnctehse wcairsclcea,lctuhlaeted. PRO9D7,6098 PROD321 ssohtuviagbethr-uusestrs,aIbcTentietthrhuhceoFnelsefesoi.g,nlrti.ettTaguhdhhtm2oeeit6ofhboo,ei\nvrevgtWeurhahpreelmuoruasosftibadedtv\atue-oareWucrklftusesoptaee:rhfoireot6\"ct-f0Woesut4tlnshasoatuteer(uiirbWatrs"oa--lacftiaiotna-. PRO4D4,6007 PRO3D1,2062 28,0PR2O5D,013PRO2D8,2061PRO2D9,201P5RO2D5,2067PRO2D5,4023PRO5D0,4P09R4O2D5,5066 user" upsursoebrc-esssutsabi-tnsustastatun/scneausp)m,pe2ba6er2erd(o7ifn2%p8r4)o,%-69o8f PRO9D3,2053 PRO3D3,2036 PRO4D9,4055 cpeesrspirno(ps6rdt9oaud%ncu)ctc,e.tssI,n7h7tFa6hsiegb.(er62ae78tn%i,otc).haielscFduodalraerftkienetdhirtoeslye PRO3D1,6004 PRO3D9,5042 PRO2D4,7076 PR1O5D66,060 color oabfovae cirtchlee, tahveerahgiegheorf th4e5%. value oRfetlahteivrealytiroa.reAusslaiggehtofosvuebr-ustsaetus Fig.F2ig8.26\.WWaaiitt--uussee"susbu-bst-asttuastbuys pbryodpurcotdsucts. of thec\an be indicated for products:b5e66
Wait-user" status can (21%), 13 (19%), 267 (21%). noted for products 604 (\Wait-user" sub-status appeared in 84% of process instances),Fo2r6S2T(s7,2c%ou)n,tr6ie9s,8fu(n6c9t%ion),daivnisdio7ns7,6lin(e67n%um).beFrsorantdhaocsteiopnroowdnuecrstst,hethsqeuraaretio is de ngriatpehlys waberoevegetnhereataevdetroageaesiolyf 4vi5s%ua.lizAe hreolwatmivaenlyy trimarees uthsee Woafitt-huese\r Wsuabi-tst-autsuesr" sub-stawtaussucsaedn abnedinwdhiactapteerdcefnotr opfroacdtuivcittises56it6a(c2co1u%nt)s, 1fo3r.(1T9h%e)s,izaenodf 2th6e7 s(q2u1a%re).
represents the frequency of sub-status usage. The color corresponds to Wait-user subSupp ostrattustecoaumnts/a.llSthTesevtehnatst cuosuenttrhaetio\.Wait-user" sub-status the most frequently are presented in Fig. 29. In this square diagram, the size of a square represents
Support teoafmtsh. eIn\FWiga.i2t7-,uSseTrs"tshuatbu-ssteaWtuasit-uusseer. sTubh-estactuoslotrhecmororsetsfpreoqnudenstlytoare the frepqreuseenntceyd. The sub-status was executed 704 times by G97. In case of this ST, the higthhe ratio bneutmwbeeernofthsueb-nsutamtubseursaogfe \isWjuasittif-iuesdebry" tshuebla-srgtaetnuusmubseersofapnedrfothrmeetdoatcatlivnituiemsb–er of the Aecvceenpttesd./Wait-user activity accounts only for 9,43% of all the activities performed by
Thtehi\s Wsuappito-rut stearm".sTuhbe-sotvaetru-ussehaosf tbheesntateuxsemcuigthetdbe70su4spteimcteeds fboyr GG9927a.nTdhSe43h–igh numbeArccoefpt\eWd/Waiati-t-uusseerr"ascutivbi-tsytaatcucosuuntssesfoirs 1j6u,s5t5i%edandby15t,h2e%laorfgaellntuhmebacetrivoitfiesactivitiespeprfeorrfmoremdbeydtbhoyseGt7ea.mHcoowrreevsepro,ndthinegl\yA.Pcacretipatlleyd, /thWeahiigt-huersethr"anaacvteirvaigteyuascacgoeuonfts only fotthhreo9sWe.4aS3iT%t-susaoerfer apsluelbrft-oshrteamtueasdctiisinvjituthsieetisf1iseptdseurbpfyoprothmrteelifdnaecb,tyit.heGa.,t9wa7lal.ittihneg afcotrivaintieusseprerrfeosrpmonedsebiys
Thceosnusipdperoerdtatseaa mnosrmGa9l 2beahnadvioSr4f3ormthaey 1bstelisnues.pTehcetesudbo-sftaatbususisinragretlhyeu\seWd abiyt-Gu9s6er" sub-sta(3t%us):atnhdeS4\2A(c4c,1e%pt)e.dT/hWisfaaictt-uisswero"rthacntoitviciitnyg aacscGo9u6natnsdreSs4p2eacltsoivpeelyrfofromr a1l6l.t5h5e% st and 15a.c2t%ivitoiefsailnl tthhee1acstuivppitoiretslipne.rfIonrcmaseedobfythetsheoSseTst,etahmis .leTadhsetohitghheecrontchlaunsioanvethraatge the two teams contribute to a better overall performance of the process. use of the \Wait-user" sub-status is partially justi ed by the fact that all the activities performed by those STs are performed on the 1st support line, i.e., waiting for a user response is considered as a normal behavior on the 1st line. The \Wait-user" sub-status is rarely used by G96 (3%) and S42 (4.1%), which is surprising as GF9i6g 2a7nd.WS4a2it-user sub-status by support teams also perform all the activiCoutnietsriinest.heIn1stFsiugp.po2rt8 linaen.d Fig. 29 the
In case of these STs, this distrib uletaiodns toofthteheconWclauisti-ounsethratsub-status usage amongthethtewoctoeaumntsriceosntriisbutperetosented. Fig. 28 presentasbtehtetercoovmerpaallripseornforomf atnhcee Accepted Waituser stoafttuhse cporoucnets/sa.ll the activities count ratio.
The same information is represented using colors tCihnoeutnhnuetmrmibeeasr.p oifTnh\FWeigari.att-2iuo9se.roV"f aluFeFisgi g.v22is97i..bW\lWeaaiitint--uusseerr"susbub-s-tsatatutussbbyyssuuppppoorrtttteeaammss. uFsiga.ge2.ss9Ttuahbti-eunsstdsa/uittcbuha-setseittsaotuttuhasdCsleedoniusufmnrmueotsbqrsetieuedrfesrn.oemqcfoIuyansectntotifFvlfryiieogtqsii.fneuusSbeS2p-nw8setetrldayactenuonidnsu,nPtForilygan.isFdi2pga9rnr2aed8tsit.eohIWnnebtdayeiidtca-o.uiusnInentFsrcuiieabgss-.esbt3yao0tpfu.rsSoT/wdahulelce-etsvent Swededne,nPaonldanPdoladannisdt,riiIbtnuisdtijiouans.tiIonfedcthabesyetWheailat-rugwseeenrduesmnubbe-rstoaftuasctiuvsiatigees performed in those and Pocolaunndtrieits. Hisaomwjueosvntegirf,yIetnhddeiabccyoauntnhbtereieslsuasrpgiseectpenrduesmfoernbtetehrde.oofvFeiraguc.steiv2o8iftetshepseurbf-osrtamtueds diune those countriteosh.iHghorwateivopevrrae,lsuIeenn.dtsIinathIcneadnciao,bm\eApascruciesspoptneecdot/efWdthafeiot-rAuctshceree"potaevcdteirvWuitasieeits-oafcctohuentseudbf-osrtamtuosredue to high ratthioanv1a4u%e. oIfnusaIelnlrdtshitaea,tauAcscticcveoitpuitenestd/ap/lWelrfatohirtem-uaesdcetriivniatctiehtisivsicctoioeuusnnttarcryac.toiTou.hnitsedis ffoarr amboorvee tthhaen 14% of all thaveearacgtiev. itiesThpeerfsoarmmeedini nfotrhmisatcioonunitsry.reTphriesseisntfeadr aubsoinvge the avarage. colors in the map in Fig. 29. Values visible in Fig. 29 indicate the frequency of sub-status usage. The sub-status is used most frequently in Fig 28. Wait-use sub-status/ all event Sweden, Poland and India. In case of Sweden ratio by countries by products and Poland it is justifyed by the large number of activites performed in those countries. However, India can be suspected for the overuse of the sub-status due to high ratio vaue. In India, Accepted/Wait-user activities accounted for more than 14% of all the activities performed in this country. This is far above the avarage.
FigFi2g9..3W0.a\iWt-uaiste-ursesru"bs-usbta-sttuastuussuasgeebbyycocuonutrnietsr.ies Function divisions. In Fig. 30, the usage of the Wait-user sub-status is analyzed with the reference to function divisions. None of the function divisions significantly distinguishes from the others. D_1 function division with the highest usage percentage (10,82%) is still close to the average. Moreover, D_1 performs the
Fig 29. Wait-user sub-status usage by countries Function divisions. In Fig. 30, the usage of the Wait-user sub-status is analyzed with the reference to function divisions. None of the function divisions significantly distinguishes from the others. D_1 function division with the highest usage percentage (10,82%) is still close to the average. Moreover, D_1 performs the 0,0624
justified.
use of the \Wait-user" sub-status is vjuissitoifnise.d.None of the function divisions signi cantly distinguishes itselft fro m2E4_85
V3_2 the others. The 1D930 1 function divi0,0624
A2_1 803 0,0805 sion, with the highest use percentage
A2_2 (10.82%), is still close to the average0,0.688
180 Moreover, D 1 performs the majority 803 0,0805 0,0853 st amnaajloyrzietyd owf iitthwroergkasr din ttohefu1nctsiuopnpodrit- line. Thus, the higher usage of the status is Thus, the higher use of the \Wait
function divisions. of it works in the 1st support line.FigF3i0g..Wa3i1t-.use\rWsuabi-ts-tuastuesr"usasguebb-sytfautnucstiounsdeivibsyions
Fig 30. Wait-user sub-status usage by function divisions Support lines. When compariunsgagetsheamsuonpgpotrhte litnherese, tshueppodrtistlriinbeusti o1st n of sub-status 3119 0,06774
E_5 D_1 A2_4
Support lines. W248 hen compari1n61 g the s94upport lines, the distribution of sub-status user" status is justi ed for D 1. 0,0853 0,1082 0,0556 2nd 873 0,05278 rd usages
among the three Support lines. When comparing sufpolploowrts thleinenevsision1sted use of the sub-status.
A2_23 119 A2_3 E_4 E_8
The majority of 01s,8006u808,0b67-74statuses a5p4pea3r1 in 15the follows the envisioned use of the sub-status. 0,0475 support lines, the distrsitbution of
1 support lines. The value of the rA2a_5tion is The majority of sub-statuses appear in therd
\Wait-user" sub-status uhisgehser afomr tohneg3 line – 7,7%. Still, this ratio st Fig 30. Wait-user sub-status usage by function divisions 1 suptphoertthlirneeess.uTphpeorvtallinueesofofliltsohvweesrrytachtlioeosneentoi-sthe avarege 6,5%. higher vfoisriotnheed3 usleinoef –th7e,7\%W. aSitti-lul,sethr"is sruabti-o is verysctlaotsues t(ocft.heFiagv.a3r2e)g,ei.6e,.5, %th.e majority Actionpoorwtlnineer.sT.hAecvtailoune oof wtheenthreaermstios,pMe7cu.i7at%lhizu,e, in performing actions in the
The majority of sub-statuses appear in the
st of sub-statuses appear in mthoere1osftftesnupth-an avar0a,06g774e (Fig. 32). All of follows the envisioned use of the sub-status. usages among the threeBrescuhpt,pEomrtil alnindeNs ina1st use Wait-user status 3119 st 1 support line.
Brecht,isEsthmigihlearnfdorNtihnea 3ursde lWineaibt-uustusperperomsrttaaliitnnuess.
1
more ochflfoigtsehenetrotfhotarhntehaeavv3aerdrraaliggneee, –(iF.7ei.,g7,.%6.3.52S%)t.i.llA,tlhlisorfatSiiebeol them spisecviearlyizceloisne ptoerthfoeramvainreggeac6t,i5o%n.s in the Fig 31. Wait-user sub-status usage by
Brecht Katia Krzysztof 0,02F81igf.un3ct2io.n \dWiviasii0ot,155-n9ussbeyr"supspuobr-t0,s13l8t1ianetunsumu0,b0s477eer by 173,0 75,0 62,0 56,0 function divisions by support line.
Support lines. When comApcatiroing otwhneerssu. pApoctritonlinoews,entehrse Mdiusthriub, ution of sub-status 3rd 225 0,07729 2nd 873 Fig 301,05.278Wait-user sub-status usage by
3rd functio2n25 divisions by support line number st majority of it works in the 1
support line. Thus, the higher usage of the status is justified.
st majority of itA2_1works in the 1 supE_1p0ort line. Thus, the higher usage of the status is Pawel 123,0 0,1330 Muthu 80,0 0,1874
Siebel 173,0 0,0281 Pawel 123,0 0,1330 0,1874 383 0,0B846 PI Challenge 2013
25
E_10 383 0,0846 D_1 161 0,1082 A2_3 54 0,0475
E_4 31 A2_5
A2_4 94
E_8
A2_2 15 180 0,0688
D_1 161 0,1082 A2_3 54 0,0475 2nd 873 0,05278 0,07729 3rd 225
A2_4 94 0,0556 E_4 E_8 31 15 A2_5 nostrmal behavior. 1 support line.
Action owners. Action oweners
them specialize in performing actions in the
WaiBtre-chutser status Process owners. Process owners Muthu, Brecht, Emil0,,1014and Nina u5E5m,0sile t0,07h7294Ne8in,0a \W4Ra8a,f0alit0,1708 0,1649 0,1011 Krzysztof user" statu173s,0 more often than t75,h0e averaPgawele pr6o2,0cess own5e6,0r (Fig. 33), but as all of more offte0n,0281than avarage (Fig.0,135592). All 1o23,0f 0,1381 Fig 31. W0a,04i7t7-user sub-status usage by 0,1330 Fredrik them specialize in performing actions on thfeun1ctsitonsduivpi69s,p0ioonrstbylisnupep,orittlinies nruamtbheerr a 0,1179 Katia Marcin 71,0 0,1014 Fredrik 69,0 0,1179
Brecht 75,0 0,1559 Marcin Fredrik 69,0 0,1179 0,1162
Emil 55,0 0,1708 Katia 62,0 0,1381
Andreas Nina 63,0 Rafal
0,1162 48,0 0,1649
48,0 Krzys0z,t1o0f11 56,0 0,0477 Fig 32. Wait-user sub-status usage by action owner
Olga 47,0 0,1237
Tomasz 44,0 0,0976 71,0 Organizat0i,1o014ns. The grapEmhil presNeinanted Raifanl Fig. 33 uses information about the
55,0 48,0 48,0 organizations concerninOlgga 0t,1h708e freq0,1u6T4o9meansz cy0,101o1f using sub-status and the information
Andreas 47,0 44,0 conce0r,1n162ing what percen0,1t237 of activit0i,0e97s6 Accept/Wait-user activity accounts for. Two
63,0 Fig 3M2uthu. Wait-user sub-sAntdreaastus usage b47y,0 action 44o,0 wner
Olga Tomasz 80,0 63,0 0,1237 0,0976 Organizations. The graph presented in Fig. 33 uses information about the organizations concerniFnigg 3t2h.eWfariet-quuseerncsuyb-ostfatuussiunsgagesubby-astcatitouns oawnnder the information
Fig. 33. \Wait-user" sub-status use by action owner. concerning what percent of activities Accept/Wait-user activity accounts for. Two Organizations. The graph presented in Fig. 33 uses information about the organizations concerning the frequency of using sub-status and the information concerning what percent of activities Accept/Wait-user activity accounts for. Two Organizations. The use of the \Wait-user" status by various organizations is visualized in Fig. 34. Organizations are laid out according to the frequency of their use of the \Wait-user" status on the x-axis, and the percentage of their activities the \Accept/Wait-user" activity accounts for.
The di erence between organizations A2, B, and C and the other organizations is due to the large volume of activities performed in general by A2, B, and C. However, two other organizations have to be distinguished: I and V1 (on the toorgpanoifzatthioen isgupraert)i.cuTlahromseusot rbgeadniisztaigtuioisnhsedp:eIrafonrdmV1a. Tlihmoisteeodrgnaunmizabteiornosfpearcftoirvmities alinmditeadlmnousmt boenreo-f ftahctiovfititehsesaendacatlimviotsitesoinse-tfhifeth\ Aofccaecptitv/itWiesaitis-uAsecrc"epat/cWtiaviitt-yu.ser activity.
0w,a012i02468t/event 0,1V84118
0,202I00 27C83 0,0660 Fig 33. Wait-user status usage by organizations
Fig. 34. \Wait-user" status use by organizations. 2.4 Process Conformity per Organization
The conformity of organizations with the assumed process model provided in [
The gure does not include 538 (13%) process instances that are absolutely unique and do not follow the model. Those are process instances involving multiple activities and are characterized by a long duration of execution.
Fig. LLL. Mainstream behavior in organization C
Fig. LLL. Mainstream behavior in organization C
The similar diagram for organization A2 is presented in Fig. FFF. In total, Fig. HHH.
organization A2 executes 4306 activities within 553 process instances. Diagram in
proceMssositnsftreaqnuceenstlyinvporolvceinssg iAns2taniscepsrienscelnudteedacintivFitiiegs. A36cc.e7pt4edv-a2nrdi,anCtosmopnfldepterdo-c2ends.s
The majority of process instances are executed by organization A2 in the 2 line. execuTthiooissnt ishfracelqveueaernblytelyevnispirdboliescionvFeinirgset.daFnFfcoeFsr. Mtinhciislssuidsnuegbaascdethitveiorteifenspcerotocepsussihntsotnadfrnocnetsms.tprlaetteegdy-2insd.the
M cess Accepted-2 , Co MTsoehsriitsopusrcoplceroeasrblslyeimvnissftiobarlneocirengsaFnaiigrz.eaFteiFoxFne.cAMu2ties–sdinthbgeyamdohraegjoraernnitciyzeaottfoiaopcnutsivAhit2tioeisfnraotrnhetensto2rtantpdeegrlyfionirsem.theMed oinst is st frequstehenreitol1yusplpirnroecb.elTesmhseifnodsriftofaernrgeacnneciszeaiitnniocntlhuAids2eas–apctehtcietvmibteaitejwosrei\etyAnocthfceeacpttwtiveoidtio-e2rsgnaadrne"iznaaotnitodpnes\rfCiosromvmiespdiblielneteind2nd".tFhiTeg. st h1HisHliinHse.c.lTehaerldyifvfeirseinbclee iinn thFiisga.s3p6ec.t between the two organizations is visible in Fig. FFF. Mainstream behavior in organization A2
Fig. FFF. Mainstream behavior in organization A2
Fig. 36. Mainstream behavior in organization A2.
80% of activities are performed by organization C in the 1st line. In case of organization A2 it is less that 40%. Missing adherence to push to front strategy is a serious problemMoforer tohragta8n0iz%atoiofnaaAct2iv.iTtiehse idsipeerrfeonrmceedinbtyhoisrgaasnpiezcattiobnetCweinenthe first line. In case A2 and C is visibleofinorFgiagn.iz3a7t.ion A2 it is less that 40%. In this asects organization C conforms to the
Fig. 36 does ennvoitsioinnecdlumdoede1l1o3f operation to the higher extent. process instances that are absolute unique and explain the lower e ciency of organization A2 operation.
In case of organization A2 those exceptional process instances account for 21% of all the process instances.
This is about 8% more than for organization C.
Concluding, organization C is more in line with mFodige.l HofHoHp.eAracttiiovnitsies performed by organizations C and A2 in various support lines than organization A2. Organization tFioign.s 3C7.anAdctAiv2itiinesvpareirofuorsmsuedppboyrtorlignaensi.zaA2 forwards too muchFiogf.itFsFoFpedroaetsionot include 113 process instances that are absolute unique and to the 2nd and 3rdcolinnter,ibaunted tthheeloewxeecruetfifoincieonfctyheofporrogcaensiszaintisotnanAc2esoipsemratuiochn.leInsscase of organization standardized leadinAg2ttohotshee ehxicgehptpioenrcaelnptraogceesosf iunsntiaqnuceesexaecccuoutinotnfso.r 21% of all the process instances.
Concluding, organization C is more in line with model of operation than 4 Tools organization A2. Organization A2 forwards too much of its operation to the 2nd and 3rd line, and the execution of the process instances is much less standardized leading 4.1 Data mungintogthe high percentage of unique executions.
3.1 Data munging The basic tool for data munging used for our analyses has been Microsoft Excel 2010. Its capacities to import CSV le have allow the original datasets to be imported to a tab3ularTofoorlms at. Next, ltering and sorting capabilities of MS Excel have eased the identi cation of missing data and have allowed for raw estimations of the count and distribution of the data, such as the number of organization lines. 4.2 Data processinTghe basic tool for data munging has been Microsoft Excel 2010. Its capacities to An important tooli mfoproortuCrSaVnafliylesehsaivsetahleloPwytthheonorliagningauladgaetaasnetds ttohebeEnimthpoourgtehdt to a tabular format. Canopy environmeNnte.xPt,yftilhtoerni,n ginaintds vsoerrtsiinogn c2a.p7a.4b,ilhitaiess boefeMnuSsEedxcteol phraovceeessastehdethe identification of cleaned data to obmtaisinsinag bdeattatearndunhdaevrestaallnodwiendg foofr rtahwe ecsotnimceaptitonosf olifntehse count and distribution of the and the relation between lindeastaa,ndsuch as the number of organization lines. The
process owners/organization lines/functions. ThEexdcaetla has been used for presented in Tabs.g3e,n4e,ra5ti,oannodf 6cohluamvenbdeiaegnracmomspinusteecdtiionn P2.y1t.hon.
The following Python modules have been useful for our analyses: { the csv module 3h.a2s bDeaetna purseodcetsosini mgport and export data from and to CSV les, { the networkx moduAlen himaspobreteannt utsoeodl ftoor corueartaenaalnydseasniaslythzee gPryatphhosn, elasnpgecuiaaglley and the Enthought the cycle graph uCsaendotpoy steundviyropnimnge-npt.onPgy.thon, in its version 2.7.4, has been used to process the cleaned data to obtain a better understanding of the concept of lines and the relation The EnthoughtbCeatwneoepny lhinaessbaenedn cpherossoenns/aosragnaneinzavtiiroonnmlineenst/fsuunpcptioorntsi.nTghoef-dthatea- presented in Tables box the networkx module and encompassing the IPython environment providing
support for Python nTotheebfooollkosw. ing Python modules have been useful for our analyses:
For visualization of control ow in both Incident and Problem Handling processes, Disco (http://fluxicon.com/disco/) has been used in version 1.3.6. Disco is a commercial software developed by the Fluxicon company. The academic license allows analysis of event logs up to 1 million events.
The process map generator (cf. Fig. 1), build-in ltering mechanism, import of CSV les and export to MXML and XES formats compatible with ProM 5 and ProM 6 applications has been used in our analyses. Disco's built-in ltering algorithm allows including or excluding process instances and activities based on the appearance of one or more properties. Although Disco does not provide support for process conformance checking, the ltering algorithms can be used to ll this gap. This is especially useful when testing various hypotheses.
The ProM Framework (http://www.promtools.org/prom6/) is an open source application for process mining. ProM perfectly complements the functionality of Disco. ProM 6 has been used to perform social network analysis (handover of work relations were examined). Dotted chart diagram (e.g., Fig. 2b) have been generated with ProM 5.
The visualization of chord diagrams is based on the D3.js javascript library, available at http://d3js.org/, and the examples of chord diagrams published by Mike Bostock (http://bl.ocks.org/mbostock/4062006 and http://bl. ocks.org/mbostock/1046712).
The visualization of the cycle graph has been performed in Gephi 0.8.2-beta (http://www.gephi.org/). The processing of betweenness and degree centralities has been performed with Gephi, as well as the laying out process and the nal rendition of the graph visualization.
Square (e.g., Fig. 33), circle (e.g., Figs. 28 and 34) and map diagrams (e.g., Fig. 30) were generated using Tableau. Tableau is a commercial software (15 day trial is available) used for visualization of data (http://www.tableausoftware. com/). The software signi cantly improves the process of nding patterns in analyzed data by providing the possibility of fast and convenient switching among various data visualizations.
The Third International Business Process Intelligence Challenge is an interesting opportunity to check the potential use of various available tools to dig into reallife data. The size of the datasets was relatively small. As a consequence, all our analyses have been performed on personal computers and do not require heavy data processing infrastructure. Analyses of the full dataset of the VINST system event logs require probably di erent tools than those used to perform the analyses described in this paper.
The datasets provided by Volvo IT Belgium is especially interesting because of the complex organizational structures (STs, organization lines, support lines, function lines) that provide support for the Incident and Problem handling processes. We have chosen to address this aspect of the datasets with the toolkit established by the social network analysis (SNA) perspective community. It seems to us that the existing tools, such as the Python networkx module and Gephi, are able to provide real insights to complex social settings. However, expertise in SNA is a strong requirement for these tools that requires an important human involvement to extract interesting knowledge from the datasets.
Process mining and SNA have shown to be complementary: process mining tools have provided insights about the dynamic character of the Incident and Process Handling processes, while SNA tools have provided insights about the social/organizational structures in which these processes are embedded. With the intensi cation of research e orts on temporal networks, SNA may with no doubt bene t from the results of the works of the process mining community.
The image of the implementation of Incident and Problem Handling processes by Volvo based on our analyses is rather positive: issues are usually localized to given STs, products, or organization lines. The push to front strategy is rather well respected, with the expectation of the 4th and 5th lines that we have identi ed during our analyses. Ping-pong behavior is rather marginal. The mis-use of the \Wait-User" sub-status is restricted mainly to few organizations. Finally, although organization A2 conforms less with the proposed process model than the organization C, most process instances are close to the proposed model.
Among future works, it would be interesting to check if our results are conrmed by larger VINST event logs.
Acknowledgments. This work has been partially supported by the Polish National Science Center. Grant no. DEC-2011/01/N/ST6/04205.