USING INFORMATION OF AN INFORMAL NETWORK

TO EVALUATE BUSINESS PROCESS ROBUSTNESS

Malika Grim-Yefsah

1,3

, Camille Rosenthal-Sabroux

and Virginie Thion-Goasdoué

LAMSADE, Paris Dauphine University, Paris, France

CEDRIC, CNAM, Paris, France

INSERM, Paris, France

Keywords: Business process, Quality evaluation, Robustness, Quality metrics, Tacit knowledge.

Abstract: We consider the evaluation of a business process quality, in particular the evaluation of its robustness. By

robustness, we mean robustness w.r.t. the risk of loosing knowledge of persons implied in the business

process. We define metrics taking tacit knowledge into account. These metrics are based on the analysis of a

social network underlying the process execution. We illustrate these metrics on a –real– application case:

the evaluation of an IS project management business process.

1 INTRODUCTION

T. H. Davenport and J. E. Short (1990) defined a

business process as being a set of logically related

tasks performed to achieve a business outcome. In

order to run efficiently, a company must identify and

manage its processes. Managing a business process

includes, among other things, monitoring its quality.

Monitoring quality means defining quality metrics

for different quality dimensions and then monitoring

them by measuring them periodically. Our goal is to

define quality metrics for the robustness quality

dimension of a business process, where robustness

measures the risk of loosing knowledge necessary to

the business process execution.

We consider here business processes consisting

of tasks performed by persons. To achieve a task, an

“official” executor often informally asks for help to

other persons that we call contributors. Help

consists in giving an advice, reminding a technical

procedure, giving an informal validation, etc. Thus,

executing a task requires not only executors’

knowledge but also contributors’ one, and more

particularly their tacit knowledge. One of the

peculiarities of tacit knowledge is that it is not

entirely “explicitable”. Consequently, the whole tacit

knowledge of a person cannot be transmitted to

another person or a system: tacit knowledge is

inherent to a person. The underlying problem here is

that if a person implied in a task execution is

missing then this task can be in peril as the adequate

-eventually tacit- knowledge required for the task

execution is missing. In this context, it is important

to be able to evaluate the robustness (of the business

processes) w.r.t. the risk of loosing knowledge

(including the tacit one).

Another research domain focuses on persons

(and implicitly on their knowledge): the social

network analysis domain, addressed by sociologists.

Social network analysis consists in (1) modeling a

social network, usually seen as a graph and (2)

analyzing this graph in order to identify e.g. social

positions, friendship groups, or central nodes. We

greatly inspire of this domain to define the notion of

informal network representing informal relations

created between persons during the execution of a

business process.

We propose quality metrics, measuring business

process robustness, linked to the presence or absence

of persons, and the risk of loosing knowledge with

regard to the informal network. We illustrate our

approach on a real application: the transition phase

of an outsourced project management business

process in a French Public Scientific and

Technological Institution (PSTI).

This article is organized as follows. We first

present our application case (in Section 2), as we use

it to illustrate the following concepts. In Section 3

we briefly discuss the concept of tacit knowledge for

business process execution. We then introduce in

Section 4 the notion of informal network underlying

430

Grim-Yefsah M., Rosenthal-Sabroux C. and Thion-Goasdoué V..

USING INFORMATION OF AN INFORMAL NETWORK TO EVALUATE BUSINESS PROCESS ROBUSTNESS.

DOI: 10.5220/0003638204300435

In Proceedings of the International Conference on Knowledge Management and Information Sharing (RDBPM-2011), pages 430-435

ISBN: 978-989-8425-81-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

a business process. Section 5 is devoted to the

definition of metrics for the evaluation of business

process robustness using informations of the

informal network. We also discuss results of the

evaluation for our application in this section. After a

presentation of related works in Section 6, we draw

conclusions and give perspectives in Section 7.

2 APPLICATION CASE

Outsourcing information system development has

become a common practice in companies. An

outsourced project usually implicates three

participants: two internal participants which are the

IS Department and the business direction concerned

by the project, and an external participant which is a

software and computing services company also

called service provider. The service provider is

chosen at the end of an invitation to tender. In a

French public organization like a PSTI, government

contract rules concerning outsourcing requires to

(re)call for tenders on a contract at least each three

years, leading to change the service provider during

the project. This change necessitates performing a

transition from the outgoing provider to the

incoming one. Here stands our application case. We

are in contact with a project manager of outsourced

project in a PSTI, which describes the transition

process as follows. The transition consists of six

activities: (Activity 1) the initialization activity

which marks the official start of the transition phase;

(Activity 2) the Third Party Maintenance (TPM)

ending where an inventory of internal and external

documents and codes is performed; (Activity 3) the

edition and validation of the transfer plan; (Activity

4) the knowledge transfer essentially consisting in

transmitting documentations, applications and codes

from the outgoing team to the incoming one;

(Activity 5) the maintenance in cooperation during

which outgoing and incoming service providers

assume together a maintenance of the application;

and (Activity 6) the responsibilities transfer, which

marks the official departure of the outgoing

provider.

A rather complex diagram formalizes the

transition. We will detail our reasoning for only two

activities of the process: activities 2 and 3, restricted

to the PSTI actor. Figure 1 presents this part of the

transition process formalized with an UML activity

diagram (the language chosen for business process

description does not matter in the following).

Figure 1: Part of the transition process.

3 BUSINESS PROCESSES AND

TACIT KNOWKEDGE

We agree with the vision considering that

(Hypothesis 1) knowledge is not an object. This

vision, explained in details by Grundstein (2009), is

based on the theories that deal with the construction

of tacit individual knowledge (Nonaka and

Takeuchi, 1995). Tacit knowledge refers to

intangible elements, inherent to the individuals who

bear them, like skills, crafts, “job secrets”, historical

and contextual knowledge, environmental

knowledge like clients, competitors, technologies,

socio-economic factors, etc. Tacit knowledge

generally cannot be entirely expressed (“explicited”

is a more recognized term of the KM community).

This means that two persons, in some cases, are not

interchangeable for a task execution in a business

process. Moreover (Hypothesis 2) a person

executing a task often informally appeals to other

persons whose (tacit) knowledge helps to a better

execution of the task. These persons do not appear in

the modeled business process, seen as the “official

procedure” in the following. Our contribution is

based on Hypotheses 1 and 2. We define metrics for

measuring the robustness, w.r.t. the risk of loosing

persons’ knowledge, in order to identify the more

sensitive tasks and activities.

Intuitively, a business process is robust if its

tasks are not in peril. A task is in peril if a part of the

knowledge needed for its execution is missing,

meaning that a person executing the task is absent or

that a person informally needed is absent. This leads

us to introduce the concept of informal network,

which can be seen as a specification of social

network for which the exchanged resource is

informal help in order to execute tasks of a business

process. As we are convinced that the major part of

the informal exchanges between employees in an

organization does not only pass through digital

supports, we consider an informal network

USING INFORMATION OF AN INFORMAL NETWORK TO EVALUATE BUSINESS PROCESS ROBUSTNESS

431

accounting for informal exchanges independently

from the communication support. Discovering the

structure of such a network rests on a questionnaire

survey of the employees susceptible to belong to the

network. Results of this survey form a network

represented by a graph, which can be analyzed. This

approach is called social network structural analysis

(Degenne and Forsé, 1999). We inspire from this

domain to define the concept of interdependencies

system permitting to model and analyze the informal

network underlying a business process.

4 INFORMAL NETWORK

We now turn to the definition of an interdependencies

system, which is a simplified version of a social

network graph. Let Tasks be a finite set of tasks and

let Persons be a finite set of persons.

Definition (Interdependencies System). An

interdependencies system is a directed graph

S=(Persons,Tasks,R) where R

⊆

Persons

Tasks

Persons

is a set of labeled directed edges.

The set Persons contains persons (executors and

contributors) implied in one of the tasks of Tasks.

For each t, we note r

the relation referring to the set

of transitions of the form (p

,t,p

), with

}

⊆

Persons, noted r

) in the following. Each

denotes help requests between persons in order to

achieve t. Intuitively, a directed edge from a person

to a person p

labeled with task t means that p

needs the informal help of p

in order to achieve t.

Definition (Interdependencies System restricted

to a Task). We note S

where t

∈

Tasks, the graph S

restricted to the relation r

For our application case, Tasks is the set of tasks

appearing in the transition process. Figure 2 shows

the interdependencies systems restricted to the

Inventory task (of course, persons’ names were

anonymized). By definition, the Inventory task

labels each edge. Henry (project manager) is an

executor of the task. He is responsible of making the

inventory of the elements manipulated in the TPM.

At is own initiative, Henry informally asks for

validation or completion of the inventory to Mola,

expert of the applicative architecture; Marion expert

of the software architecture, who herself informally

asks for help to Arnold (database administrator),

Sallah (front office), and Elsa (JAVA developer);

René, expert of the hardware architecture, who

himself informally asks for help to Walter and

George (system and network engineers); and Marcus

functional contact, who himself informally asks for

help to three business experts: human resources

(Charles), application (Irina) and scientific (Indiana).

It is important to note that options and alternatives

cannot be expressed in an interdependencies system.

(we one cannot express that a person is optional for

the execution of a task or that a person can substitute

another). In order to be as specific as possible, we

consider that if r

(p,p

) for all i in [1..n] then every p

where i in [1..n] is necessary to p in order to achieve t.

Figure 2: S

|Inventory

(S restricted to the Inventory task).

For the illustration of our application case, we

also present the interdependencies system restricted

to the Ending validation task in Figure 3.

Figure 3: S

|Ending

validation (S restricted to the Ending

validation task).

Definition (Accessibility). The person p’ is

accessible from the person p for a task t in the

interdependencies system S, noted Needs(S,p,p’,t),

iff there is a path form p to p’ in S

For our application (see Figures 2 and 3), one has

for example Needs(S, Marcus, Charles, Inventory),

Needs(S, Henry, George, Inventory), Needs(S,

Henry, Karen, Ending validation) and also Needs(S,

Henry, Steven, Ending validation).

5 ROBUSTNESS EVALUATION

Conceptual representations like business processes

usually come from a conception phase based on

needs analysis. In this kind of representations, actors

are described at the role level. This suggests that

two persons having the same role are

interchangeable. In practice, this hypothesis is

approximate (see the complete discussion in Section

3). In our application case for example, if Marcus

needs the help of Charles (HR expert) in order to

achieve a task then another HR expert usually

cannot replace Charles without negative impact on

the task execution quality. In other words, Charles’s

KMIS 2011 - International Conference on Knowledge Management and Information Sharing

432

knowledge is the only knowledge that helps Marcus

as efficiently as possible. An informal network

identifies which person needs informal help from

another. These two visions are complementary but

the level of granularity in an interdependencies

system is the person while it is the role in the

business process. Thus, it is necessary to map

persons and tasks in order to map the informal

network and the business process. Let’s note

Execute (p,t), where p

∈

Persons and t

∈

Tasks be

the relation denoting this information (p executes

task t). This relation, instanciated thanks to the

interviews, intrinsically maps interdependancies

system and business process.

For our application case, we identified several

executors: Henry (project manager), Earl

(administrative assistant), Wu (administrator) and

Wilhelmina (project co-manager), with Execute

(Henry,Inventory); Execute (Henry,Order), Execute

(Earl, Order), Execute (Wu, Order), and Execute

(Wilhelmina, Order) meaning that Henry, Wu, Earl

and Wilhelmina are co-executors of the task;

Execute

(Henry,Edition); Execute (Henry,Ending validation)

and Execute (Wilhelmina, Ending validation);

Execute (Henry,Reversibilty validation).

One has to note the fundamental distinction

between an executor and a contributor. An executor

appears in the official procedure associated to the

task. For example, according to the official

procedure, Henry is an executor of the Ending

validation task. The official procedure also stipulates

that Henry must ask for Wilhelmina’s validation for

the execution of this task. In this context,

Wilhelmina is also an executor of the task.

Contributors are Steven and Karen (see Figure 3)

who are persons that an executor (Henry) informally

ask for help to. Wilhelmina does not appear in

Figure 3 this shows that she does not ask for help to

anyone. Steven and Karen are the only contributors

for this task.

5.1 Definition and Measurement of

Metrics for Robustness Evaluation

We present here some metrics, at the task level.

Definition (Metric “Global Sensitivity of a

Task”). For a task, this metric counts the number of

persons implied in the task: executors plus

contributors (that appear in the interdependencies

system). The higher is the measure, the riskier is the

task. For a task t, this metric, noted

global_sensitivity(t) is defined by Cardinality(I),

where I is the set defined by

{p’

∈

Persons | Execute(p’,t) or there is p

∈

Persons

such that (Execute(p,t) and Needs(S,p,p’,t))}.

For instance, for the Ending Validation task (see

Figure 3 and the instanciation of the Execute

relation), one has I={Henry, Wilhelmina, Karen,

Steven}, so global_sensitivity(Ending validation) =4.

For the Inventory task (see Figure 2 and the

instanciation of the Execute relation), one has

global_sensitivity(Inventory)=13.

Definition (Metric “Sensitivity by Depth of a

Task”). For a task, this metric measures the

maximal size of a path going from an executor to a

contributor. Intuitively, the larger is the path, the

riskier it is to go from an executor to a contributor (if

a person is missing then the path is “broken”). In the

following, Max(s), where s is a set of integers,

returns the higher element of s (and returns 0 if s is

empty); and Max_path(executor,contributor,S’), where

{executor, contributor}

⊆ Persons and S’ is an

interdependencies system, returns the size of the

larger path from executor to contributor in S’. For a

task t, the sensitivity by depth metric, noted

sensitivity_by_depth(t) is defined by Max(depth_paths)

where depth_paths is the following set:

For instance, for the Ending validation task (see

Figure 3), sensitivity_by_depth(Ending validation)=1.

For the Inventory task (see Figure 2),

sensitivity_by_depth(Inventory)=2.

Now, let’s consider a metric measuring the

density of the informal network underlying a task t.

The density is a well-known metric used in the

social network analysis community. It measures the

number of non oriented connections devided by the

number of possible non oriented connections

(number of non oriented connexion in the

corresponding strongly connected graph). The

highter is the measure, the denser is the network and

so the more tolerant is the network to the absence of

a person, as persons are very connected (they “know

each other”). Contrary to previous metrics,

performing the measurement with the graph

restricted to the considering task would be limitative

because, if persons know each other, that is not

necessary via this specific task execution. We then

define the density for the whole informal network S

limited to the persons (but not the relations) implied

in the task. We decide to measure a dispersion (1-

density) for uniformization with the other metrics

preserving the highter

is riskier convention for

results interpretation.

USING INFORMATION OF AN INFORMAL NETWORK TO EVALUATE BUSINESS PROCESS ROBUSTNESS

433

Definition (Metric “Dispersion of the Informal

Network underlying a Task”). For a task t, this

metric noted dispersion(t) is defined by:

where |Persons

| is the number of vertices from S

and |E

| is the number of pairs of the set E

defined

by { {p

} | p

∈

Persons

and p

∈

Persons

and

) or r

) with t

∈

Tasks) } i.e. the set of

pairs of persons, implied in the task t, connected by

a task -any task- of Tasks.

As an activity or a business process consists of a

set of tasks, metrics for business process robustness

can be defined by aggregation of the robustness

metrics of its tasks (e.g. by sum, average, maximum,

weighting tasks metrics according to the task

importance, etc., eventually taking decision nodes

into account). By lake of place, we do not consider

such metrics here.

Table 1 presents the results of metrics evaluation

for the tasks of activities 2 and 3. These measures

show that the Inventory task is more sensitive than

the majority of the other tasks. Indeed it implies a

large number of persons (global_sensibility) with a

high dispersion of the network (dispersion), meaning

there are lots of persons implied in the Inventory

task and they are poorly connected together. One can

observe the same phenomenon for the Edition task,

which additionally presents a longer path executor-

contributor (sensitivity_by_depth) than the other tasks.

Table 1: Measurement of metrics on the application case.

global_

sensitivity

sensitivity_

by_depth

dispersion

Activity 2 - TPM Ending

Task Command

7 2 0,5

Task Inventory

0,8

Task Ending validation

4 1 0,3

Task Edition

12 3 0,8

Activity 3 - Edition and valid. of the transition plan

Task Transition valid.

4 2 0,5

Metrics presented here allowed identifying

sensitive “zones” (activities and tasks) of a business

process, this identification being explained by

objective measures. The study points the TPM

ending activity as being the most sensitive one of the

business process. Within this activity, two tasks

were noticed particularly sensitive: Inventory and

Edition ones. We can draw several conclusions: 1-

these tasks are more complex to achieve than we

thought before the study, the executors seeking for a

lot of informal help (besides the official procedure),

2- the absence of persons (not appearing in the

official procedure), could negatively impact the

execution quality of these tasks.

Based on the results on this study, we can

consider several business perspectives:

(1) Monitoring of the sensitive tasks execution

quality. As a rule, a very special attention has to be

paid to the execution quality of sensitive tasks and

activities, even more particularly if a situation can

create departure or moving of contributors (e.g.

reorganization of the entity a contributor belongs to,

or more simply to tasks performed during summer

vacation periods).

(2) Improvement of the procedures. Informal

contributors whose knowledge is absolutely

necessary to a task should appear in the official

procedure (adding new official “sub-tasks”).

Nevertheless, the precision level of the business

process description is delicate to find. Indeed a very

precise procedure insures a best execution of the

process but often slows its execution. Furthermore a

very complex procedure is often hardly accepted

because it is more difficult to execute and can make

the job “off-putting”.

6 RELATED WORK

Quality metrics of interest were proposed for

business processes (Vanderfeesten et al., 2007). In

particular, lots of contributions concern the

complexity metric, which can be seen as a factor for

the understandability dimension. One can also cite

the cyclomatic number (McCabe, 1976) (Gruhn and

Laue, 2006), (Cardoso et al., 2006), the Conrol-Flow

Complexity (Cardoso, 2008), or the size (Cardoso et

al., 2006) (Gruhn and Laue, 2006). The robustness

factor is well-studied in the multi-criteria decision

aiding domain (see (Aissi and Roy, 2009) for an

overview) but, as far as we know, not in control flow

oriented processes. Another close work is (Hassan,

2009). In this work, N. Hassan measures IT-enabled

business process performance by evaluating the

impact of an IS evolution (i.e. the implementation of

a new technology). Considering the IS system as an

IT actor, he analyses a social network of IT actors

before and after the implementation of the new

technology. Based on this analysis, he draw

conclusions concerning for instance the adoption of

the new IS system or the evolution of the business

job.

Concerning the knowledge facet, we point that

none of all these works explicitly consider persons

KMIS 2011 - International Conference on Knowledge Management and Information Sharing

434

and their tacit knowledge implied in the business

process execution. We believe that our method

brings a complementary vision by focusing on

persons’ tacit knowledge.

7 CONCLUSIONS AND

PERSPECTIVES

In this article, we deal with the quality evaluation of

a business process, in particular its robustness

evaluation. We propose metrics for the evaluation of

–a part of– the robustness using information coming

from the analysis of an informal network (a social

network for which the resource is informal help).

We illustrate metrics on a real application case: the

transition phase of an outsourced project

management in a French PSTI. We also discuss

some business perspectives based on results of this

evaluation.

The application case illustrating the metrics

constitutes a “proof of concept”. One has to note that

we could consider lots of other metrics and quality

dimensions for the definition of a business process

quality (or even just robustness). A second more

detailed study aiming at defining the quality more

comprehensively (and consequently considering

other dimensions and metrics) on the same

application case is in progress. For this second

study, we use the GQM paradigm (Basili, Gianluigi,

and Rombach, 1994) that gives a methodology for

quality requirements elicitation.

Metrics definition can be improved through

different ways. (1) Metrics are defined in function of

the system of interdependencies only. We could go

further by using results of the analysis of the system

-performed by sociologists- in order to define other

metrics. We think about stability of the network (the

stabler is the network, the easier a person can find

help through it), centrality of persons (if there is a

central person, communication between persons is

facilitated), or similarities of persons (if a person p

is absent, a similar person p

could eventually

replace p

, minimizing the impact of the absence of

). (2) As discussed in Section 4, the formalism

used to model the interdependencies system does not

permit to express alternatives or options. Another

more expressive formalism should be considered

(e.g. and/or graphs) in order to enrich the

interdependencies system thus refine the definition

of metrics. (3) Social network analysis domain deals

with a notion of resource, which is what a person

needs from another one. Thus, a network is not

simply a graph, but a set of graph, one for each

resource. For our application, the only resource to be

observed between persons was the need of informal

help. It would be interesting to characterize different

resources in order to express more astute metrics.

REFERENCES

Aissi, H. and Roy, B., (2009). Robustness in Multi-

Criteria Decision Aiding, Book chapter. New Trends

in Multiple Criteria Decision Analysis, Springer

Science + Business Media.

Basili, V., Gianluigi, C. and Rombach, D. H. (1994). The

Goal Question Metric Approach, Encyclopedia of

Software Engineering, J. Wiley & Sons, 528-532.

Cardoso, J. (2008) Business Process Control-Flow

Complexity: Metric, Evaluation, and Validation. Int.

Journal of Web Services Research. 5(2), 49-76.

Cardoso, J., Mendling, J., Neumann, G. and Reijers, H. A.

(2006). A Discourse on Complexity of Process

Models, In Proc. of the Intl. Conf. on Business Process

Management. LNCS 4103, Springer, 117-128.

Davenport, T. H. and Short, J. E. (1990). The New

Industrial Engineering: Information Technology and

Business Process Redesign. Sloan Management

Review. 11–27.

Degenne A. and Forsé M. (1999). Introducing Social

Networks, Sage Publications.

Gruhn, V. and Laue, R. (2006). Complexity metrics for

business process models, In Proc. of the 9th Intl. Conf.

on Business Information Systems, LNI P-85, 1-12.

Hassan, N. (2009). Using Social Network Analysis to

Measure IT-Enabled Business Process Performance,

Information Systems Management, 26 (1), 61-76.

McCabe, T. J. (1976) A Complexity Measure, IEEE

Transactions on Software Engineering, 308-320.

Nonaka, I. and Takeuchi, H. (1995). The knowledge-

Creating Company: How Japanese Companies Create

the Dynamics of Innovation, Oxford University Press.

Vanderfeesten, I., Cardoso, J., Mendling, J., Reijers, H. A.

and van der Aalst, W. (2007). Quality Metrics for

Business Process Models, In Proc. of BPM and

Workflow Handbook. Future Strategies, 179-190.

USING INFORMATION OF AN INFORMAL NETWORK TO EVALUATE BUSINESS PROCESS ROBUSTNESS

435