De ising an SLA-Awa e Me hodology o
Imp o e P ocess Pe o mance?
Ca los M¨ulle ?, Minsu Cho, Pablo Fe nandez?, Minseok Song, Manuel
Resinas?, and An onio Ruiz-Co ´es?
?Uni e si y o Se ille, Uni e sidad de Se illa, Spain.
Pohang Uni e si y o Science and Technology, Pohang, Ko ea.
Abs ac . Aiming o be as compe i i e as possible, o ganisa ions a e
always pu suing o imp o e hei business p ocesses applying co ec i e
ac ions when needed. Howe e , he ac ual analysis and decision mak-
ing o hose ac ions is ypically a challenging ask elying on ex ensi e
human-in- he-loop expe ise. Speci ically, his imp o emen p ocess usu-
ally in ol es: (i) o analyse e idences o unde s and he cu en beha -
io ; (ii) o decide he ac ual objec i es (usually de ined in Se ice Le el
Ag eemen s -SLAs- based on in ui ion) and (iii) o es ablish he im-
p o emen plan. In his ongoing wo k, we aim o p opose a da a-d i en
and in ui ion- ee me hodology o de ine an SLA as a go e nance ele-
men ha speci ies he se ice le el objec i es in an explici way. Such a
me hodology conside s p ocess pe o mance indica o s ha a e analysed
by means o in e ence, op imiza ion, and simula ion echniques. In o de
o mo i a e and exempli y ou wo k we add ess a Heal hca e scena io.
1 In oduc ion o ou P oposed Me hodology
Nowadays, in a changing and compe i i e business wo ld, o ganisa ions mus
keep hei business p ocesses (BPs) unde con ol. In o he wo ds, hey mus
analyse i cu en BPs beha iou is as expec ed o no . Howe e , such an anal-
ysis is no easy o be add essed and au oma ed because i elies on ex ensi e
human-in- he-loop expe ise in se e al asks, namely: (i) o ex ac in o ma ion
om e idences (such as logs o incidences) o unde s and he cu en beha iou ;
(ii) o decide, based on in ui ion, he ac ual objec i es in Se ice Le el Ag ee-
men s (SLAs) ha go e n he o ganisa ion; and (iii) o es ablish he imp o e-
men ac ions. In his ongoing pape we p opose a da a-d i en me hodology ha
au oma ically ga he s an SLA o go e n he o ganisa ion ac i i ies by means o
Se ice Le el Objec i es (SLOs). As one o i s ad an ages, ou p oposal does no
equi es ha a p e ious SLA we e ini ially de ined by o ganisa ions. To he con-
a y, being da a-d i en, i c ea es an SLA jus om a se o e en logs om he
?This wo k was pa ially suppo ed by he Eu opean Commission (FEDER), he Eu-
opean Union Ho izon 2020 esea ch and inno a ion p og amme unde he Ma ie
Sklodowska-Cu ie g an ag eemen No 645751 (RISE BPM), he Spanish and he
Andalusian R&D&I p og ams (g an s TIN2015-70560-R, P12-TIC-1867), he Na-
ional Resea ch Founda ion o Ko ea (No.NRF-2014K1A3A7A030737007).
cu en BPs beha iou ; and a se o P ocess Pe o mance Indica o s (PPIs) [1, 5].
A ollowing we in oduce he me hodology (c. . Fig.1) using as unning example
he p ocesses o a Hospi al In . Sys em (HIS) wi hou a p ede ined SLA.
P ocess
Mining
SLA
De ini ion
Simula ion P ocess
Mining’
Op imizing
SLOs
In e ing
SLOs E alua ion
PPIs
E en
Logs
CSLOs DSLOs
E en
Logs’
PPIs’
SLA
Manage s
Fig. 1. SLA-Awa e Me hodology o Imp o e O ganisa ions Pe o mance.
Ini ially, we assume ha he HIS gene a es a numbe o e en logs including
di e en da a (e.g. ini ial and end momen , s a e, e c [6]) as esul o ope a ing
on mul iple p ocesses (e.g. egula o eme gency oom p ocesses, e c) o pe o m
se e al heal hca e ac i i ies (e.g. magne ic esonance imaging -MRI-, X- ay, e c).
In a p ocess mining ac i i y he e en logs a e p ocessed in conjunc ion wi h
a numbe o PPIs de ined by hospi al manage s (e.g. ”P obabili y ha wo king
ime o pa ien s who ake MRI es is less han 71 min.”). As esul o he p o-
cess mining we would ge he measu es o all PPIs and hey a e he inpu o
an in e ing SLOs ac i i y. Such an ac i i y in e s he Cu en Se ice Le el
Objec i es (CSLOs in Fig. 1) o selec ed, manually o no , PPIs. These CSLOs
deno e he ac ual cu en HIS beha iou o he PPIs (e.g. ”P obabili y ha
wo king ime o pa ien s who ake MRI es is less han 71 min. is mo e han
95%”) conside ing some o ganisa ional ac o s in ol ed in he Hospi al BPs (e.g.
”wi h 15 employees and 3000$ pe employee.”). A his poin , he CSLOs a e
p ocessed by some op imiza ion echniques in an op imizing SLOs ac i i y
ha ga he s a numbe o desi ed SLOs (DSLOs in Fig. 1) deno ing he desi ed
beha iou ha Hospi al manage s would like o achie e. The unde lying op i-
miza ion echnique [4, 2] can be con igu ed by Hospi al manage s conside ing he
men ioned o ganisa ional ac o s. An example o DSLO could be ”P obabili y
ha wo king ime o pa ien s who ake MRI es is less han 71 min. is mo e
han 99%”) achie ed by speci ic alues o he o ganisa ional ac o s (e.g. ”wi h
17 employees and 2700$ pe employee.”). A e ha , we p opose o analyse i he
applica ion o DSLOs esul s in a p ope BPs beha iou o no . In doing so, a
simula ion model gene a ed om he ini ial e en logs [3] can be modi ied based
on DSLOs, hen a new se o e en logs (e en logs’ in Fig. 1) is ob ained and
i is inpu ed in o a new p ocess mining’ ac i i y o ge he simula ed mea-
su es o all PPIs (PPIs’ in Fig. 1). Then, an e alua ion ac i i y is pe o med o
analyse he de ia ion be ween simula ed measu es o PPIs’ and he DSLOs. In
case o de ia ion some ac i i ies a e execu ed again bu wi h di e en se ings
(o ganisa ional ac o s o DSLOs) om he Hospi al manage s. Speci ically, in
case o high de ia ion he in e ing SLOs ac i i y is execu ed again; and in
case o low de ia ion, he op imizing SLOs ac i i y is execu ed again. Finally,
i he de ia ion is negligible, he SLA de ini ion ac i i y is pe o med o ga he
he documen ha will go e n he BPs in u u e.
2 Conclusions and Fu u e Wo k
In his ongoing wo k we de ise an SLA-awa e and in ui ion- ee me hodology o
imp o e o ganisa ions pe o mance ha ga he s an SLA o go e n o ganisa ions.
The main bene i s o ou me hodology a e he ollowing: (i) he SLA is ga he ed
om ac ual BPs beha iou and no om sc a ch as adi ionally based on man-
age s in ui ion; (ii) he imp o emen is assu ed by simula ing and e alua ing he
p ocesses beha iou be o e c ea ing he SLA; and inally, (iii) he manual in e -
ac ion in he whole p ocess is limi ed o some manual se ings (c. . Fig.2 ha
depic s ou ision o a u u e ool o help manage s in he DSLOs de ini ion).
PPI 1
PPI 2
PPI 3
PPI 4
PPI 5
PPI1
“P obabili y ha wo king ime o pa ien s
who ake MRI es < 71.0 min. is mo e han 95%”
-15 employees
-3000$ pe employee
- Cu en SLO
“P obabili y ha wo king ime o pa ien s who
ake MRI es < is mo e han %”
- Desi ed SLO
95%
71.0
99%
71.0
- 1+ employee o imp o emen s Calcula e
Weigh
71
PPI2
GLOBAL
Op imiza ion
Hi e employees
Incen i e o $.
- Incen i e o 100$ o imp o emen s
Simula ion
DSLO 1 
DSLO 2 
DSLO 5 
PPI 3
PPI 4
99
Fig. 2. Mockup o he desi ed con ol panel.
Re e ences
1. A. del-R´ıo-O ega, M. Resinas, C. Cabanillas, and A. Ruiz-Co ´es. On he de ini ion
and design- ime analysis o p ocess pe o mance indica o s. In . Sys ., 38(4), 2013.
2. C. M¨ulle , M. Resinas, and A. Ruiz-Co ´es. Au oma ed Analysis o Con lic s in
WS–Ag eemen Documen s. T ansac ions on Se ices Compu ing, 2013.
3. A. Rozina , R.S. Mans, M. Song, and W.M.P. an de Aals . Disco e ing simula ion
models. In o ma ion Sys ems, 34(3):305–327, 2009.
4. A. Ruiz-Co ´es, O. Ma ´ın-D´ıaz, A. Du ´an, and M. To o. Imp o ing he Au o-
ma ic P ocu emen o Web Se ices using Cons ain P og amming. In . Jou nal
on Coope a i e In o ma ion Sys ems, 14(4), 2005.
5. W.M.P. an de Aals . P ocess cubes: Slicing, dicing, olling up and d illing down
e en da a o p ocess mining. In Asia Paci ic Business P ocess Managemen - Fi s
Asia Paci ic Con e ence, 2013, Beijing, China, 2013. Selec ed Pape s, 2013.
6. S. Yoo, M. Cho, E. Kim, S. Kim, Y. Sim, D. Yoo, H. Hwang, and M. Song. As-
sessmen o hospi al p ocesses using a p ocess mining echnique: Ou pa ien p ocess
analysis a a e ia y hospi al. In . Jou nal o Medical In o ma ics, 88, 2016.