BUSINESS PROCESS PRIORISATION

WITH MULTICRITERIA METHODS

Case of Business Process Reengineering

Elena Kornyshova and

Camille Salinesi

University Paris 1 - Panthéon Sorbonne, 90, rue de Tolbiac, 75013 Paris, France

Saint-Petersburg State University of Economics and Finance, 21, Sadovaia Str, 191023 Saint-Petersburg, Russia

Keywords: Business Processes Prioritisation, Multicriteria Methods, Business Process Reengineering.

Abstract: Business process (BP) engineering is used nowadays in many methods, techniques and tools. In domains such

as strategic management, reengineering, or security analysis, one particular concern is the identification of BPs

that should be dealt primarily. In practice, the number of BPs is often very large and it justifies the creation of a

priorisation mechanism. However, the number of approaches available to prioritise BPs specifically is very

limited. This paper presents a comparison of multicriteria (MC) methods, and an approach to guide the

selection and application of the MC method found as the most appropriate for BP priorisation. The approach is

illustrated with the case of selecting and applying a BP priorisation in the view of BP reengineering.

1 INTRODUCTION

Many BP engineering methods, techniques and tools

focus on “Key business processes” (Sachdeva and

Joshy, 2005) (PegaRules, 2003). BPs priorisation is

used by companies to define the most important

development axes, to increase the reaction speed to

environment changes, to optimize the expenditure,

and consequently, to improve their competitiveness.

Dealing with Key BPs supposes that decision

makers know business processes priorities or are able

to define them, at least intuitively. The intuitive

approach is viable when there is a limited number of

BPs. However in most cases, managers face problems

with a large number and large variability of BPs, and

often different versions of BPs through time. The

combination of these issues leads to a combinatory

explosion of the number of artefacts to deal with,

hence a better priorisation support is needed.

There is a limited number of researches dealing

with BPs priorisation. In our point of view, BP

priorisation can be considered as a multicriteria

decision problem, and therefore we suggest integrating

multicriteria (MC) methods into BP priorisation. Our

aim is to propose a formal approach for BP priorisation

in order to enhance decision-making (DM) in the field

of BP management and related fields such as system

engineering, or business security.

We develop our approach to achieve two main

goals: (i) selecting an appropriate MC method and (ii)

applying it to the considered BP priorisation case.

We suggest that a process allowing to guide the

selection of a DM method should take into account the

multiple aspects of the situation at hand. The presented

approach copes with these different aspects using a

structured benchmarking grid. The grid was adapted

from (Papadacci et al., 2006) to the MC methods

comparison issue and is applied to describe BP

priorisation problem, which includes the description of

alternative BPs and criteria typology for BP evaluation.

The rest of the paper is organised as follows:

section 2 gives an overview of existing approaches of

BP priorisation; section 3 presents MC approach for

defining BP priorities and justifies the selection of one

MC method; section 4 illustrates our approach with

example of BPR. The section 5 discusses possible

application domains and research's perspectives.

2 OVERVIEW OF EXISTING BP

PRIORISATION APPROACHES

This section presents an overview of existing BPs

priorisation approaches. After a brief description of

these approaches, we compare them and give some

conclusions.

138

Kornyshova E. and Salinesi C. (2007).

BUSINESS PROCESS PRIORISATION WITH MULTICRITERIA METHODS - Case of Business Process Reengineering.

In Proceedings of the Ninth International Conference on Enterprise Information Systems - ISAS, pages 138-143

DOI: 10.5220/0002381901380143

 SciTePress

There is only a small number of approaches that

propose to guide BP selection. Four approaches are

particularly considered in our review: (i) Hammer and

Champy’s, (ii) Robson and Ullah’s, (iii) PROSCI, and

(iv) Mazur's et al. approaches.

Hammer and Champy (Hammer and Champy,

1993) propose to analyse BPs under three different

perspectives in order to select those that need

reengineering. These are "problems", "importance"

and "feasibility". First, all processes for which a

problem can be identified are chosen. Then, the

importance of these BPs for the organization is

analysed. Last, a feasibility control is carried out in

order to verify if expected results will cover related

expenses.

Robson and Ullah (Robson and Ullah, 1996)

propose a methodology to sort BPs for reengineering.

In this approach, BPs are analysed in relation with

critical success factors (CSFs). First, relevant CSFs

are listed, and then each BP is estimated along a five-

grade scale according to all CSFs. A weighted sum is

generated for each process; weights represent relative

importance of CSF. It presents a complex value of

each BP for organization. Besides, the authors

suggest to analyse BP functioning (from very good to

bad according to five-grade scale). Three levels of BP

priorities are finally considered: reengineering,

improvement and supervision. BPs that contribute to

many CSFs and have bad functioning are considered

as potential for BP reengineering.

PROSCI (Crowe et al., 1997) uses a BP taxonomy

to identify reengineering opportunities. The authors

suggest that relations exist between strategic goals

and BPs. The first step of this approach consists in

establishing the taxonomy of BPs. In the second step,

the influence of each BP on every strategic goal is

taken into account. To achieve this, an influence

diagram is drawn using decision tree where every BP

is embedded in a main decision node, strategic goals

are drawn as chance nodes, and main decision nodes

are linked to each chance node. Relationships are in

the form of probability distributions, which reflect the

stochastic nature of influences that BPs have on the

strategic goals. The final BP evaluations are obtained

using weighted sums of chance nodes in which

weights are assigned to chance nodes depending on

their order of importance.

Mazur et al. (Mazur et al., 2000) propose an

approach for BP selection based on weighed sum. In

this approach, the calculation is made according to the

next criteria: influence on customer, variability,

functioning, and importance for business. Each BP is

measured towards all criteria with five-grade scale,

and then the weighted sum is calculated.

The four selected approaches are compared along

two dimensions: (i) the criteria used by the

approaches for comparing BPs and (ii) the rules

proposed to carry out BP selection. Several remarks

can be made: (i) there is only a limited set of criteria

to support BP comparison; (ii) most criteria are

abstract (e.g. problems, or importance for customers

and business), and the authors do not show how these

criteria relate to actual BP performance indicators;

and (iii) there are only two kinds of selection rules,

weighted sum and two-dimensional space. The

drawback of weighted sum is that it requires

homogeneous criteria. On the other hand, two-

dimensional space has the disadvantage of limiting

maximal number of criteria.

In order to avoid these issues, we suggest

integrating MC methods into BP priorisation.

3 MULTICRITERIA

PRIORISATION OF BP

As indicated above, our proposal consists in using

multicriteria methods in order to carry out BP

priorisation. MC methods are very different from each

other, and the result of priorisation highly depends on

the selected method. We believe that a MC method

must take into account the specific characteristics of

problems situation to provide appropriate results.

Therefore, we propose an approach that guides the

selection of a MC method consistent with the

situation at hand.

As Fig. 1 shows it, the guidance provided by our

approach is based on a 5 phases process. The process

results in applying a MC method specifically chosen

to deal with the problem at hand.

1. Define the

MultiCriteria

Problem

2. Identify

Candidate MC

Methods

4. Select a

Method

3. Evaluate

Candidate Methods

against the Problem

5. Apply the

Selected Method

Figure 1: Overview of the proposed approach.

The goal of the initiation phase is to define the

nature of the MC problem. Once the problem defined,

it is necessary to identify candidate methods (phase

2), to evaluate their ability to cope with the MC

problem (phase 3), and to select the most adequate

method(s) (phase 4). Phases 2, 3 and 4 are iterative as

several phases can match the problem at hand (in

which case a more detailed analysis is required) or on

BUSINESS PROCESS PRIORISATION WITH MULTICRITERIA METHODS - Case of Business Process Reengineering

139

the contrary none of the candidate methods matches

the problem perfectly. In this case, another cycle of

evaluation must be achieved. Several strategies are

available: either other methods are considered, or

some of the required characteristics are added or

removed, or the characteristics are ranked by order of

importance. For more details, see (Salinesi and

Kornyshova, 2006).

3.1 Details of the Approach in the

Context of BP Priorisation

Our approach provides different ways of working

depending on the actual problem to deal with. The

following sections develop the approach application

in the specific context of BP priorisation.

3.1.1 Defining Multi-criteria Problem

Based on a state of the art research (Papadacci et al.,

2006), we developed a benchmarking grid that helps

defining a MC problem in detail. The grid is

composed of 15 different facets organized into four

orthogonal dimensions, namely: context, process,

form, and object.

The context dimension gathers 5 characteristics of

the situation of method use: (i) the problem is a choice

(ii) ranking, (iii) or sorting, (iv) new alternatives can

emerge, and (v) there are multiple viewpoints.

The process dimension gathers 4 characteristics of

the expected way of method applying: (i) the

approach for defining evaluations (either unique

criterion of synthesis (UCS), or outranking), (ii) for

defining the decision criteria (either without

weighting, with weighting and interdependencies, or

simple weighting), (iii) the ability to deal with

different measure scales, and (iv) easiness of use

(easy, medium or difficult).

The form dimension characterizes how the

method is described. This dimension gathers two

parameters: (i) notation (textual explanation,

mathematical formula, function), and (ii) tool (to

indicate if a software support is available).

The object dimension describes the alternatives to

be prioritized using 4 characteristics: (i) type of data

to consider (either quantitative or qualitative), (ii)

number of alternatives that will be considered with

the method (either large or small), (iii) ability to take

into account incompatibilities and conflicts between

alternatives, and (iv) hierarchicality (ability to deal

with alternatives organized within a hierarchy tree).

In the context of BP priorisation, the problem can

be a choice (application example is BPR), a ranking

(for example, the BPs must be ranked in order to

establish priorities for business security

improvement), and sorting (e.g. BPs are positioned

according to the Capability Maturity Model).

In our case, the potential actions to be considered

are the BPs of an organisation. Their number may

vary from little (if only “macroprocesses” are taken

into account) to very large (if all detailed BPs of the

BP hierarchy are considered). It is very important to

take into account the hierarchical nature of BPs.

Indeed (i) only BP of the same hierarchical level

should be compared, and (ii) BP analysis must taken

into account the nature of the hierarchical links

between BPs. Besides, the BP collection is dynamic.

New BPs emerge, some disappear, and some change

their properties. Alternative BPs have various nature

and may be evaluated according to multiple criteria.

We suggest the 9 following criteria drawn from

literature (Voyer, 1999), (PegaRules, 2003), (Shadrin,

2002), (Sachdeva and Joshy, 2005) and (Crowe et al.,

1997):

 BP duration,

 BP quality,

 BP cost,

 BP size,

 BP customers satisfaction,

 BP efficiency,

 BP productivity,

 BP contribution to strategic goals,

 BP problems,

and, the 4 following criteria, developed based on

our experience with BP priorisation:

 BP contribution to problems resolution (the

number of problems that can be solved by

improving the given BP),

 BP lifecycle steps: creation, development,

stable functioning, regress, and destruction,

 BP influence on stakeholders,

 BP customer: internal or external.

These criteria have different scales: cost and value

are absolute numerical data, efficiency and

productivity are ratio, contribution to strategic

objectives, life cycle steps have nominal scales etc. In

addition, data type takes two values: quantitative and

qualitative.

Besides, the analysis involves multiple

stakeholders with different, and sometimes

contradictory, viewpoints.

This analysis allows characterizing the situation in

which BP priorisation shall be undertaken and shall

help selecting an appropriate multicriteria method.

3.1.2 Identifying Candidate MC Methods

The analysis grid was applied to the four general-

purpose MC methods: Multiattribute Utility Theory

ICEIS 2007 - International Conference on Enterprise Information Systems

140

(MAUT) (Keeney and Raiffa, 1993), Analytical

Hierarchy Process (AHP) (Saaty, 1980), Outranking

methods (Roy and Bouyssous, 1993), and Weighting

methods (Keeney, 1999). For the sake of space, these

methods are not detailed here. However, table 1

shows an overview using the benchmarking grid.

Table 1: Overview of considered MC methods.

Dimension

Facets

MAUT AHP

Outran-

king

Weigh-

ting

Context

Problematic, choice Yes Yes Yes Yes

Problematic, ranking Yes Yes Yes Yes

Problematic, sorting No No Yes No

Treatment of a new

alternative

Yes No Yes Yes

Taking into account

multiple viewpoints

No No Yes No

Process

Approach for defining

evaluations

UCS UCS Out-

ranking

UCS

Approach for decision

criteria weighting

Yes, no

interd

Yes,

interd

Yes,

interd

Yes, no

interd

Taking into account

various scales of criteria

Yes No Yes No

Easiness of use Difficult Easy Medium Easy

Form

Notation Utility

func-

tion

Weigh-

ted sum

Textual Weigh-

ted sum

Tools No Yes Yes Yes

Object

Data type quan,

qual

quan,

qual

quan,

qual

quan

Number of alternatives to

be treated

Great Small Great Great

Treatment of

incompatibility,

alternatives conflicts

Yes No Yes No

Hierarchicality No Yes No No

3.1.3 Evaluating Candidate Methods

The goal of this step is to identify which candidate

method satisfies all the characteristics which are at

step 1.

In our example: (i) all the considered methods

deal with the choice and ranking problems, and only

outranking methods allow alternatives sorting; (ii)

two methods (MAUT and outranking) supports

various scales of criteria and deal with a great

alternatives number; (iii) AHP is not able to treat the

apparition of new alternatives, and (iv) only

Outranking is able to deal with multiple viewpoints.

3.1.4 Selecting and Applying a Method

Both MAUT and outranking methods satisfy majority

of characteristics. Nevertheless, outranking methods

exceed MAUT regarding to two criteria: sorting

problematic and ability to take into account the

multiple viewpoints. If the last criteria are not

significant, then two methods are equivalent. In such

a case, we must extend analysis to other criteria.

Besides characteristics elicited on step 1, these

methods differ according to approach for defining

evaluations, easiness of use and tool presence. The

approach by outranking gives a more exact result then

unique criterion synthesis. Moreover, outranking

methods are more easy to use and are supported by

tools.

Thus, our recommendation is to use outranking

methods for defining BP priorities.

4 CASE STUDY WITH ELECTRE

This section presents a case study undertaken at a

company in the electronics industry. The purpose of

the experiment was to choose BP to be reengineered.

As a result of this experience, the enterprise expected

to identify one or two processes which reengineering

would bring maximal value with minimal drawbacks.

As shown in the former section, an outranking

method should be considered to deal with this issue.

The family of ELECTRE methods was found

particularly interesting by the enterprise. The

ELECTRE I method intended for choice problems

(Roy and Bouyssous, 1993) was finally chosen. In

order to apply ELECTRE I, one must (i) define the

problem (potential BP and criteria) and evaluate BP

according to selected criteria, and (ii) apply the

method.

4.1 Problem Definition

The problem definition includes specifying (i) a list of

BP, (ii) a list of criteria, (iii) criteria construction and

preference rules, and (iv) criteria weights.

The set of BPs was developed based on the

"Process classification framework" proposed by

APQC (Process Classification Framework, 1996). It

included:

 BP1.Understand Markets and Customers,

 BP2.Design Products,

 BP3.Market and Sell,

 BP4.Produce and Deliver,

 BP5.Invoice and Service Customers,

 BP6.Develop and Manage Human Resources,

BUSINESS PROCESS PRIORISATION WITH MULTICRITERIA METHODS - Case of Business Process Reengineering

141

 BP7.Manage Information Resources,

 BP8.Manage Financial and Physical Resources.

The collection of criteria to be considered while

applying ELECTRE I was defined based on the

enterprise requirements:

Cr.1. BP contribution to strategic goals presents

the degree of influence of BPs on organizational

performance. Within the framework of Balanced

Scorecard, strategic goals are divided into two

categories: "results" that concern financial

performance and customers, and "leverages" that

concern internal processes, learning and growth.

Weights were distributed within these categories: 2 –

"results" and 1 – "leverages". The final evaluation

was based on weighted sum. The preference rule was

maximum.

Cr.2. BP contribution to problems resolution

means that BPR should help to resolve some decision

problems. The problem at hand was to select a BP

which improvement would bring the greatest result.

We defined improvement by the contribution of

processes to problems that could be solved by process

reengineering. To make the analysis closer to reality

the frequency of occurrence and threat degree were

used as weights. For frequency, the scale was: 0 –

never, 1 – sometimes, 2 – often, 3 – regular. For threat

degree, the scale was estimated on a three-level grade:

1 – low, 2 – medium, 3 – high. This function was

maximum.

Cr.3. BP costs were defined as the number of

persons, working on the BP. Preference rule was

minimum.

Cr.4. BP sizes were defined by the quantity of

sub-processes, which we believed would reflect their

importance in the company. The preference function

was aimed at a maximum.

Cr.5. The purpose of BP life cycle steps was to

define the administrative influences required for

process reforming. Indeed, it was found that process

reengineering was needed or at least acceptable for

processes in the state of development, regression or

stable functioning. Reengineering was felt less

preferable for BPs in state of creation, destruction and

stable functioning. Therefore, the preference rule was

defined as: (development = regress) ≥ stable

functioning > (creation = destruction).

Cr.6. BP customers could be external or internal.

External processes add value for organization's

customers, therefore they were considered as more

important. The preference rule is: external ≥ internal.

To define criteria weights, we used the SWING

method (Keeney, 1999). The decision maker (DM)

chose the most important criterion and affected a

value of 100 to it. Then, the DM chose the most

important criterion and affected a lower value to it.

The same principle was applied recursively until a

value was affected to all criteria. Normalisation

produced weights as shown in Table 2.

Table 2: Criteria Weighting.

Criteria Cr.1 Cr.2 Cr.3 Cr.4 Cr.5 Cr.6

Value 80 100 20 20 60 50

Weight 0.24 0.30 0.06 0.06 0.19 0.15

The two first criteria were general; that is they

included "sub-criteria". In order to define the partial

evaluations we attributed “1” to BPs, which affected

either strategic goals (in our case, the data on strategic

goals were taken from Balanced Scorecard (BSC)) or

problems to be solved. The final evaluations are the

weighted sums of the partial ones (the weights are

described above).

We proceeded by simply assigning values to BP

evaluation according to next four criteria. The

summary of BP evaluation is presented in Table 3.

Table 3: BP evaluation summary.

Criteria

weights

BP1

BP2

BP3

BP4

BP5

BP6

BP7

BP8

Cr.1 (in points)

0,24 18 5 12 18 2 2 1 9

Cr.2 (in points)

0,30

3 4 13 13 4 4 4 26

Cr.3 (in persons)

0,06 4 8 5 29 2 1 3 2

Cr.4 (sub-processes

number)

0,06 6 1 1 5 2 2 3 4

Cr.5 (nominal)

0,19 st.

fun.

st.

fun.

st.

fun.

reg. reg. st.

fun.

cre-

at.

reg.

Cr.6 (nominal)

0,15 ext. int. ext. ext. ext. int. int. int.

4.2 ELECTRE Application

ELECTRE I is based on the principles of concordance

and discordance (see Roy and Bouyssous, 1993).

The method starts by a calculation of concordance

and discordance indices. These indices define

concordance and discordance with the assumption that

alternative A is preferred to alternative B. Concordance

and discordance were established using the following

principle: if a DM declared that alternative A is at least

as good as B for the majority of attributes then a

concordance was defined. Discordance was defined

based on the other attributes according to which A was

not strong enough compared with B. All calculations

are not shown here for the sake of space. The

concordance and discordance matrices developed in

our case study are shown in Table 4 and Table 4.

Using a threshold of 0,55 to highlight BPs in the

concordance and discordance tables revealed that BP8

ICEIS 2007 - International Conference on Enterprise Information Systems

142

(Manage Financial and Physical Resources)

dominated the others on average without any

particular shortcoming in terms of discordance. A

qualitative analysis of this choice revealed that the

enterprise agreed with it. A reengineering of the

financial and physical resources management

processes was thus undertaken.

Table 4: Concordance matrix.

BP1 BP2 BP3 BP4 BP5 BP6 BP7 BP8

BP1 0,70 0,70 0,51 0,45 0,64 0,64 0,45

BP2 0,49 0,25 0,06 0,54 0,88 0,88 0,15

BP3 0,64 1,00 0,51 0,69 0,88 0,88 0,39

BP4 0,88 0,94 0,94 0,94 0,94 0,94 0,64

BP5 0,70 0,76 0,46 0,40 1,00 0,94 0,40

BP6 0,55 0,76 0,31 0,06 0,66 0,94 0,21

BP7 0,36 0,57 0,12 0,06 0,36 0,51 0,15

BP8 0,55 1,00 0,61 0,55 0,79 0,94 1,00

Table 5: Discordance matrix.

BP1 BP2 BP3 BP4 BP5 BP6 BP7 BP8

BP1 0,04 0,43 0,43 0,14 0,14 0,04 1,00

BP2 0,76 0,39 0,76 0,25 0,25 0,14 0,96

BP3 0,35 0,00 0,35 0,18 0,18 0,07 0,57

BP4 0,86 0,75 0,82 1,00 1,00 0,89 0,93

BP5 0,94 0,18 0,59 0,94 0,00 0,20 0,96

BP6 0,94 0,18 0,59 0,94 0,50 0,20 0,96

BP7 1,00 0,24 0,65 1,00 0,11 0,11 0,96

BP8 0,53 0,00 0,18 0,53 0,07 0,07 0,00

5 CONCLUSIONS

This paper shows how to choose and to apply a

MCDM method. In the particular domain of BP

priorisation, it shows that outranking methods should

be used. Based on an analytical comparison, this

claim is confirmed by a case study of BP priorisation

for the purpose of reengineering in an electronic

company.

Besides BPR, BP priorisation could be achieved

in others contexts such as: ERP implementation,

business continuity plan elaboration, or improvement

of Information System strategic alignment. We

believe that other case studies in these domains and

comparative analyses should be undertaken to fully

validate our approach.

Defining BP priorities with a structured MC

method has advantages: (i) time for decision-making

and implementing decreases thanks to less analysis

mistakes in the BP, (ii) expenses decrease, (iii) degree

of goals achievement grows by targeting the most

important BPs, and (iv) stakeholders confidence in

results and in the overall project grows owing to their

participation in the definition of priorities.

We intend to proceed this research in two

directions: improving our approach to multicriteria

methods selection and developing new practical cases

to obtain more precise evaluations.

REFERENCES

Crowe, T.J., Rathi, K. and Rolfes, J.D., (1997), Applying a

Taxonomy of Business Processes to Identify

Reengineering Opportunities,

http://www.prosci.com/rathi.htm, October 15, 2005.

Hammer, M. and Champy, J., (1993), Le reengineering:

Réinventer l'entreprise pour une amélioration

spectaculaire de ses performances, Dunod, Paris.

Keeney, R.L., (1999), Foundations for Making Smart

Decisions, IIE Solutions, 31, No. 5.

Keeney, R.L. and Raiffa, H., (1993), Decisions with

Multiple Objectives: Preferences and Value Trade-Offs.

Cambridge University Press.

Mazur, I., Shapiro, V., Titov, S. and Elkina, L., (2000),

Companies restructuring, Moscow, Ed. High school.

Papadacci, E., Salinesi, C. and Sidler, L., (2006) Panorama

des approches d’arbitrage dans le contexte de

l'urbanisation du SI, Revue des sciences et techniques de

l'information (RSTI), Hermes, France.

PegaRULES Process Commander. Business Activity

Monitoring: Empowering BPM, (2003),

http://www.insurancenetworking.com/assets/article/4/B

AM%20White%20Paper.pdf, October 19, 2005.

Process Classification Framework (1996),

http://www.apqc.org/, February 28, 2006.

Robson, M. and Ullah, P., (1996), A Practical Guide to

Business Process Reengineering, Gower Publishing

Limited.

Roy, B. and Bouyssous, D., (1993), Aide Multicritère à la

Décision: Méthodes et Cas, Economica, Paris.

Saaty, T.L., (1980), The Analytic Hierarchy Process. NY,

McGraw Hill.

Sachdeva, N. and Joshy, J., (2005), On demand business

process life cycle, Part 8: Business process monitoring

– Create key performance indicators, http://www-

128.ibm.com/developerworks/webservices/library/ws-

odbp8/, October 19, 2005.

Salinesi, C. and Kornyshova, E. (2006), Choosing a

Prioritization Method – Case of IS Security

Improvement. In Forum Proceedings of the 18th

International Conference on Advanced Information

Systems Engineering, Luxembourg.

Shadrin, A., (2002), Process approach: Foundation and

implementation methods, All about Quality, N 16-17.

Moscow, Ed. Trek.

Voyer, P., (1999), Tableaux de bord de gestion et

indicateurs de performance, Presse Universitaire du

Québec.

BUSINESS PROCESS PRIORISATION WITH MULTICRITERIA METHODS - Case of Business Process Reengineering

143