Process Adaptation Patterns
for Cross-Organizational Business Process Modeling
Moufida Aouachria
1,2,3
, Abderrahmane Leshob
1
and Abdessamed R
´
eda Ghomari
4
1
Laboratory for Research on Technology for Ecommerce (LATECE), University of Quebec at Montreal, Montreal, Canada
2
Division T
´
elecom, Centre de D
´
eveloppement des Technologies Avanc
´
ees, Algiers, Algeria
3
Universit
´
e de Batna 02, Batna, Algeria
4
LMCS Laboratory,
´
Ecole Nationale Sup
´
erieure d’Informatique, Algiers, Algeria
Keywords:
Business Process Modeling, Business Patterns, Business Process Integration, Process Adaptation Patterns,
Process Mining, Workflow Net.
Abstract:
Nowadays organizations collaborate through cross-organizational business processes. These business pro-
cesses require the coordination of several partners who are often geographically dispersed. Modeling such
processes is complex and requires that designers have extensive experience in particular when organizations’
processes are incompatible. This paper addresses the problem of modeling cross-organization processes out
of collection of organizations private process models. To this end, we propose a set of process adaptation
patterns that connect private processes and resolve interoperability issues. Proposed patterns are formalized
with workflow net.
1 INTRODUCTION
Nowadays organizations collaborate through business
processes that cross organizations’ boundaries. These
business processes often require the coordination of
several organizations that are often geographically
dispersed. Cross-organization processes must take
into account collaborative scenarios involving dis-
tributed and autonomous partners. For example, a
typical ‘Sales & Distribution’ process requires the co-
ordination of several business partners including the
buyer, the supplier, the carrier, and other related part-
ners such as financial institutions. Modeling such pro-
cesses is complex and requires that designers have ex-
tensive experience (Zeng et al., 2013).
Coordination complexities come from the fact that
1) organizations use their business processes to col-
laborate with multiple partners who may have dif-
ferent business models (i.e., way of doing things),
2) private processes that need to be connected may
be incompatible (e.g., structural mismatch between
messages), and finally 3) these processes are often
supported by have information systems. Coordina-
tion complexities have been addressed much more at
software level (e.g., web services) through message-
based patterns (e.g., (Wang et al., 2007; Li et al.,
2010)). Such patterns are the focus of our processes
adaptation operators.
This paper proposes a set of workflow patterns
we call process adaptation patterns that aim to: 1)
help modeling cross-organizational processes, 2) re-
solve interoperability issues between partners’ pro-
cesses, and 3) improve process flexibility. Proposed
patterns are formalized with workflow net (Van der
Aalst, 1998), a formal language for modeling work-
flow processes.
The rest of the paper is organized as follows: Sec-
tion 2 introduces the problem and the motivations of
this work. Section 3 provides an overview of our ap-
proach and describes our research vision. Section 4
presents the process adaptation patterns. Section 5
surveys related work. We conclude in section 6.
2 THE RESEARCH PROBLEM
Consider two companies: Company A (Buyer) wants
to collaborate with Company B (Supplier). Company
A buys its products using the (private) procurement
process of Figure 1. Company B sells its products us-
ing the (private) Sales and Distribution (SD) process
of Figure 2.
As shown in the BPMN model of Figure 1, the
124
Aouachria, M., Leshob, A. and Ghomari, A.
Process Adaptation Patterns for Cross-Organizational Business Process Modeling.
DOI: 10.5220/0006470201240129
In Proceedings of the 14th International Joint Conference on e-Business and Telecommunications (ICETE 2017) - Volume 2: ICE-B, pages 124-129
ISBN: 978-989-758-257-8
Copyright © 2017 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

Figure 1: Procurement private process.
Figure 2: Sales and Distribution private process.
procurement process starts when it receives a pur-
chase requisition. The process then sends a request
for quotation (RFQ) to potential suppliers who, in
turn, prepare quotations and submit them back to the
company. After receiving quotations, Company A se-
lects a supplier, creates a purchase order (PO) and
sends it back to that supplier. Once the products are
received, a goods receipt is generated and the pay-
ment is made.
One of the peculiarities of this procurement pro-
cess is that the requester (i.e., Company A) does not
need to receive an invoice to generate a payment. In-
deed, it uses the Two-way match strategy for invoice
processing. Unlike, the traditional three-way match
approach consisting of matching invoice, ’PO’, and
receiving reports, Company A uses only the ’PO’ and
the goods receipt to calculate the invoice value.
On the other hand, the ’SD’ process (Figure 2) of
Company B starts by receiving an RFQ from a pur-
chaser. Company B then prepares a quotation and
sends it back to the purchaser. After receiving the
PO, Company B fulfills the order and delivers it to its
clients. Once the products are delivered, an invoice
is generated. The process ends once the payment is
received.
To enable these two organizations to collaborate,
we need to link their two private processes to build
a single new collaboration process which crosses the
organizations boundaries. This is what we call Cross-
Organizational Business Process. Unfortunately, this
process has an interoperability issue. Indeed, the sup-
plier (Company B) sends an invoice which is not ex-
pected by the requester (Company A).
This problem has two major issues. First, it makes
the new cross-organizational process invalid in the
context of this collaboration. Secondly, it will pre-
vent both processes from collaborating using infor-
mation systems that support them. Such incompati-
bility problems are the focus of our process adaptation
patterns presented in section 4. To resolve the interop-
erability issue presented in this example, we propose
a process adaptation pattern called Single-Entry-Zero-
Exit (SEZE) (see subsection 4.1) that hides the invoice
message sent by the supplier.
Before going into the details of each process adap-
tion pattern, we will present in the next section an
overview of our research vision.
3 OUR RESEARCH VISION
We would like to provide organizations with tools to
help them model cross-organization processes that ac-
curately reflect their way of doing things from a col-
lection of private process models. To this end, we
propose a three-step approach as illustrated in Figure
3. The first step analyzes the input processes (i.e., pri-
vate processes of the organizations that plan to collab-
orate) to identify interoperability issues using a pro-
cess mining approach. More precisely, we plan to an-
alyze the logs captured by the different systems that
support each private process. The second step uses
process adaptation patterns specification, we propose
in this paper, as input to identify the patterns that re-
solve interoperability issues identified in the first step.
The last step applies identified patterns to adapt input
processes (organizations’ private processes) and build
a single cross-organizational process.
The work presented in this paper deals exclusively
with the formalization of process adaptation patterns
Process Adaptation Patterns for Cross-Organizational Business Process Modeling
125

Figure 3: Overall process of the research objectives.
(see Figure 3) we derived from the service oriented
architecture (SOA) literature ( see (Wang et al., 2007;
Li et al., 2010)).
4 CROSS-ORGANIZATIONAL
PROCESS ADAPTATION
PATTERNS
The purpose of this work is to assist organizations
in the process of modeling cross-organizational
processes. This, by resolving incompatibility
issues when combining different organization’s
private processes. To this end, we have identi-
fied a set of six process adaptation patterns that
resolve message mismatches in process collabora-
tion, namely Zero-Entry-Single-Exit

ZESE
adapt

,
Single-Entry-Zero-Exit

SEZE
adapt

, Single-Entry-
Multiple-Exits

SEME
adapt

, Multiple-Entries-
Single-Exit

MESE
adapt

, Multiple-Entry-Multiple-
Exit

MEME
adapt

, and Single-Entry-Single-
Exit

SESE
adapt

. Each pattern is a process fragment
consisting of adaptation activities. For the sake of
simplicity, we choose to demonstrate our ongoing
work by formalizing two patterns we used the most
within our experimental data, namely

SEZE
adapt

and SEME
adapt
.
Next, we propose two definitions: 1) a definition
of each activity in a cross-organizational process, and
2) a definition of Adapt
WFnet
, an extended Workflow
net (Van der Aalst, 1998), we propose to model pro-
posed process adaptation patterns. We choose Work-
flow net language mainly for its formal semantic, ex-
pressiveness, graphical nature, and its ability to an-
alyze processes (Van der Aalst, 1998; Zeng et al.,
2013). It will be assumed throughout that the reader
is familiar with WF
net
(see (Van der Aalst, 1998; Zeng
et al., 2013)).
Definition 1: An activity A in cross-
organizational process is defined as an 9-
tuple(ID, ST, ET, Org
Pro
, Org
Rec
, M
s
, M
r
, R
Rel
, R
Req
),
where:
• ID, its identification. ST , its start time. ET , its
end time.
• Org
Pro
, a private organization’s process that sends
message to A.
• Org
Rec
, a private organization’s process that re-
ceives messages from A.
• M
s
=
{
M
i
, 1≤i≤n
}
, set of messages sent by A.
• M
r
=
{
M
i
, 1≤i≤m
}
, set of messages received by
A.
• R
Rel
=
{
R
i
, 1≤i≤n
}
, set of released resources af-
ter the execution of A.
• R
Req
=
{
R
i
, 1≤i≤m
}
, set of required resources
by A to execute.
To formalize our process patterns, we adapted
RM
W F
net
(Zeng et al., 2013) using the following def-
inition:
Definition 2: We call Adapt
W F
net
a Petri net com-
posed of 4-tuples (P, A; F, M
0
), where P is a finite set
of places, A a finite set of transitions, F a set of arcs ,
P∩A =
/
0, and M
0
is the initial marking of Adapt
W F
net
where:
1. P = P
L
∪ P
R
∪ P
M
, P
L
∩ P
R
=
/
0, P
L
∩ P
M
=
/
0, and P
R
∩ P
M
=
/
0; P
R
⊆ P represents the re-
sources in the workflow; and P
M
⊆ P represents
the exchanged messages in the workflow; and
P
L
represents other places which are different
from P
R
∪ P
M
.
2. F = F
L
∪ F
R
∪ F
M
, where:
(a) F
L
= (P
L
× A)∪(A ×P
L
), represents the logical
structure of the cross-organizational model;
(b) F
R
= (P
R
× A) ∪ (A × P
R
),∀x, y ∈ A ∪ P
R
,
(x, y) ∈F
R
iff (y,x)∈F
R
. F
R
, represents the re-
source relations of the cross-organizational
model;
(c) F
M
= (P
M
× A) ∪ (A × P
M
), ∀x, y ∈ A ∪ P
M
,
(x, y) ∈ F
M
iff (y, x) /∈ F
M
. F
M
, represents the
message relations of the cross-organizational
model;
ICE-B 2017 - 14th International Conference on e-Business
126

3. A=A
s
∪A
a
∪A
r
∪A
l
, where:
(a) A
s
⊆ A, represents sending activities;
(b) A
a
⊆ A, represents adaptation activities;
(c) A
r
⊆ A, represents receiving activities;
(d) A
l
= A\
{
A
s
∪A
a
∪A
r
}
, represents remaining ac-
tivities; where:
∀A
i
, A
j
∈
{
A
s
,A
a
,A
r
,A
l
}
/if A
i
6=A
j
then A
i
∩A
j
=
/
0.
4. (P
L
,A;F
L
) is a WF-net.
5. ∀ p∈P, M
0
(P) =
(
1 i f p=P
R
∪
{
O
}
0 otherwise
, Where:
O ∈ P
L
∧
•
O=
/
0. Where:
•
O =
{
y/y ∈ P∪A ∧ (y, O) ∈ F
}
is the pre-set of
O.
As shown in the definition 2 above, the tran-
sition set A in an Adapt
WF
net
is used to represent
workflow activities using four components, namely
A
s
,A
a
,A
r
, and A
l
. Compared to RM
W F
net
(see (Zeng
et al., 2013)), transition set A represents the activi-
ties in the whole cross-organizational process and not
only in the private processes.
The rest of this section describes two process
adaptation patterns using Adapt
WF
net
. For each pat-
tern we provide its name, description, an example
that illustrates the pattern, the problem, and the solu-
tion. To ensure that the resulting formalization is safe,
we adapted the mechanism proposed by Wang (Wang
et al., 2007). This mechanism allows only one work-
ing cycle of the adapter to execute at a time. In our
adapter patterns, this blocking mechanism is applied
via the usage of resource tokens. Resource tokens are
consumed once a working cycle starts, and released
when the cycle completes. Resource place is repre-
sented by a double-line circle with a token however
message place is represented by a double-line circle
without a token (Zeng et al., 2013).
4.1 Single-Entry-Zero-Exit Adaptation
Pattern

SEZE
adapt

4.1.1 Description
This adapter has the ability to delete a message sent by
one process to a target process where the latter does
not expect it.
4.1.2 Example
After receiving a purchase order, the provider process
sends a receipt while the receiving process does not
expect such message. The adapter should just hide or
delete this message after the communication ends.
4.1.3 Problem
The source process has an extra message that the tar-
get process does not expect to receive.
4.1.4 Solution
When one process sends a message that is not ex-
pected by the target process, the adapter should just
kept this message, it can be deleted after the commu-
nication ends.
Let say that a sending activity A
si
sends a message
M
si
from Org
Pro
but this message is not expected by
the receiving process (Org
Rec
). In this case, process
adaptation pattern SEZE
adapt
should be inserted to
hide this message. Therefore, SEZE
adapt
pattern is
defined by two activities
{
A
si
, A
ak
}
as follow:
(

A
si
, ST
si
, ET
si
,
/
0, Org
Rec
,
{
M
si
}
,
/
0,
/
0,

R
reqi

(A
ak
, ST
ak
, ET
ak
, Org
Pro
,
/
0,
/
0,
{
M
rk
}
,
{
R
relk
}
,
/
0)
Where:





A
si
.R
reqi
= A
ak
.R
relk
= R
i
A
si
.M
si
= A
ak
.M
rk
= M
i
A
si
.ET
si
≺ A
ak
.ST
ak
.
The Adapt
WFnet
of SEZE
adapt
pattern (See Figure
4) can be defined as a tuple (P, A; F, M
0
) where:









P =
{
M
i
}
∪
{
R
i
}
,
A =
{
A
si
}
∪
{
A
ak
}
;
F =
{
(A
si
, M
i
), (R
i
, A
si
)
}
∪
{
(M
i
, A
ak
), (A
ak
, R
i
)
}
,
M
0
= [0, 1]
Figure 4: SEZE adaptation pattern (SEZE
adapt
) .
This adapter has only a single entry sending activ-
ity A
si
got by place M
i
. The message getting from M
i
will not exit to receiving activity. Single to zero adap-
tation is performed by A
ak
. SEZE
adapt
pattern starts
a working cycle upon the run of A
si
and completes
the cycle upon the run of A
ak
. Accordingly, the run of
A
si
consumes the resource token in place R
i
and the
occurrence of A
ak
returns the token back to R
i
.
Process Adaptation Patterns for Cross-Organizational Business Process Modeling
127

4.2 Single-Entry-Multiple-Exits
Adaptation Pattern

SEME
adapt

4.2.1 Description
This adapter has the ability to split a message M
si
sent by one process while this message is expected
by the target process as N fragments of messages
M
r1
, M
r2
, .. ., M
rN
.
4.2.2 Example
When a physician writes prescription, then he sends
it within a single message, with multiple instructions
(M
r1
,M
r2
,. .. ,M
rN
) to a blood analysis laboratory.
The latter expects to receive it in multiple different
messages. The adapter must split the message into
multiple messages (M
r1
, M
r2
, .. ., M
rN
) and ensures
that all messages are supplied to the blood laboratory.
4.2.3 Problem
A process sends a single message to a target pro-
cess while the latter expects to receive this message
in fragment.
4.2.4 Solution
When a process (Org
Pro
) sends a message, with mul-
tiple information to target process (Org
Rec
) that ex-
pects it in multiple different messages; the adapter
splits the message sent by process Org
Pro
and ensures
that all messages are supplied to Org
Rec
.
Let say that an activity A
si
sends a mes-
sage M
si
from Org
Pro
but the receiving process
Org
Rec
expects to split it to receive. In this case,
process adaptation pattern SEME
adapt
should be
inserted to split the message M
si
into N message
{
M
r1
, M
r2
, .. ., M
rN
}
expected by the receiving
process (Org
Rec
). Then, SEME
adapt
pattern is
defined by activities

A
si
, A
ak,
A
r1
,A
r2
,. .. ,A
rN

as
follow:























(A
si
,ST
si
,ET
si
,
/
0,Org
Rec
,
{
M
si
}
,
/
0,
/
0,

R
req1
,R
req2
,. .. ,R
reqN

)
(A
ak
,ST
ak
,ET
ak
,Org
Pro
,Org
Rec
,
{
M
sk
}
,
{
M
rk
}
,
/
0,
/
0),
(A
r1
,ST
r1
,ET
r1
,Org
Pro
,
/
0,
/
0,
{
M
r1
}
,
{
R
rel1
}
,
/
0),
(A
r2
,ST
r2
,ET
r2
,Org
Pro
,
/
0,
/
0,
{
M
r2
}
,
{
R
rel2
}
,
/
0),
......
(A
rN
,ST
rN
,ET
rN
,Org
Pro
,
/
0,
/
0,
{
M
rN
}
,
{
R
relN
}
,
/
0).
where:











A
r j
.R
relj
= A
si
.R
req j
= R
j
, ∀ j, 1≤ j ≤ N,
A
ak
.M
sk
= ∪
∀ j, 1≤ j≤N
A
r j
.M
rj
,
A
ak
.M
rk
= A
si
.M
si
= M
si
and A
si
.ET
si
≺ A
ak
.ST
ak
,
A
ak
.ET
ak
≺ A
r j
.ST
r j
, ∀ j, 1≤ j ≤ N.
The Adapt
WFnet
of SEME
adapt
pattern (See Figure
5) can be defined as a tuple (P, A; F, M
0
) where:































P=
{
M
si
,M
r1
,M
r2
,. .. ,M
rN
}
∪
{
R
1
,R
2
,. .. ,R
N
}
,
A=
{
A
si
}
∪
{
A
ak
}
∪
{
A
r1
,A
r2
,. .. ,A
rN
}
;
F =
{
(A
si
,M
si
), (R
1
,A
si
), (R
2
,A
si
), .. ., (R
N
,A
si
)
}
∪
{
(M
si
,A
ak
), (A
ak
,M
r1
), (A
ak
,M
r2
), .. ., (A
ak
,M
rN
)
}
∪{(M
r1
,A
r1
), (M
r2
,A
r2
), ..,(M
rN
,A
rN
),
(A
r1
,R
1
), (A
r2
,R
2
), ..,(A
rN
,R
N
)}
M
0
=


0,. .. ,0
| {z }
N+1
, 1,. .. ,1
| {z }
N


.
Figure 5: SEME adaptation pattern (SEME
adapt
) .
The adapter has only a single entry sending activ-
ity A
si
got by place M
si
. The message getting from
M
si
will exit to N receiving activities. Single to Mul-
tiple adaptations from the entry to the exits is per-
formed by A
ak
. The adapter splits one message M
si
to N messages: M
r1
, M
r2
, ..., M
rN
. SEME
adapt
pat-
tern starts a working cycle upon the run of A
si
and
completes the cycle upon the run of the N receiv-
ing activities
{
A
r1
, A
r2
,. .. , A
rN
}
. Accordingly, the
run of A
si
consumes the resource tokens in places
ICE-B 2017 - 14th International Conference on e-Business
128

{
R
1
, R
2
, ..., R
N
}
and the occurrence of the N receiv-
ing activities
{
A
r1
, A
r2
,. .. , A
rN
}
returns all tokens to
Triggers
{
R
1
, R
2
, ..., R
N
}
.
5 RELATED WORK
Patterns have been introduced in the workflow area
for analyzing the expressiveness of business processes
(Van der Aalst et al., 2003). In (Weber et al., 2008),
Weber et al. proposed a set of 18 change workflow
patterns and 7 change support features to enhance
flexibility in the context of process-aware informa-
tion systems (PAIS). However, proposed patterns do
not support advanced change scenarios (e.g., adapt-
ing data when changing control).
In (Leshob et al., 2017), Leshob et al. in-
troduced a pattern-based approach to adapt generic
cross-organizational processes according to the orga-
nizations’ specific needs. While this approach adapts
business process views (i.e., dynamic, functional, or-
ganizational and informational) and insures their con-
sistency, it does not specialize/adapt information sys-
tems that support them.
In (Fdhila et al., 2015), Fdhila et al. presented
an approach to enable process adaptation in cross-
organizational processes. Authors proposed a generic
change propagation approach to adapt all partners’
processes when a partner adapts its private process.
However, proposed algorithms consider the applica-
tion of one change at a time where, in practical sce-
narios, several change operations might be applied in
a combined manner.
In (Ben et al., 2015), authors proposed six adap-
tation patterns for modeling collaborative processes.
However proposed patterns lack formalization, to ob-
tain unambiguous pattern definitions which will allow
their implementation in collaborative process support
systems.
6 CONCLUSION
Modeling cross-organizational business processes is
complex and requires that designers have extensive
experience. Indeed, modeling such processes require
putting together a collection of private business pro-
cesses from multiple partners that are often geograph-
ically dispersed. The challenge is even bigger when
private processes are incompatible or have systems
that support them such as PAIS.
The purpose of this ongoing work is to as-
sist organizations in the process of modeling cross-
organizational processes. To this end, we identified a
set of six adaptation patterns that resolve incompati-
bilities when integrating organizations’ processes.
Although this work is still at an early stage, this
paper establishes guidelines to advance our long-term
research project which consists of 1) analyzing private
processes, 2) identifying incompatibilities, 3) select-
ing process adaptation patterns, and 4) constructing
cross-organizational processes (see Section 2) using
an iterative process.
REFERENCES
Ben, S., Chaabane, M., Bouaziz, R., and Andonoff, E.
(2015). Flexibility of collaborative processes using
versions and adaptation patterns. In Research Chal-
lenges in Information Science (RCIS), 2015 IEEE 9th
International Conference on, pages 400–411. IEEE.
Fdhila, W., Indiono, C., Rinderle-Ma, S., and Reichert, M.
(2015). Dealing with change in process choreogra-
phies: Design and implementation of propagation al-
gorithms. Information systems, 49:1–24.
Leshob, A., Mili, H., Gonzalez-Huerta, J., and Boubaker, A.
(2017). A value-oriented approach to business process
specialization: Principles, proof-of-concept, and val-
idation. Journal of Systems and Software, 127:120–
149.
Li, X., Fan, Y., Madnick, S., and Sheng, Q. Z. (2010). A
pattern-based approach to protocol mediation for web
services composition. Information and Software Tech-
nology, 52(3):304–323.
Van der Aalst, W. M. (1998). The application of petri nets to
workflow management. Journal of circuits, systems,
and computers, 8(01):21–66.
Van der Aalst, W. M., Ter Hofstede, A. H., Kiepuszewski,
B., and Barros, A. P. (2003). Workflow patterns. Dis-
tributed and parallel databases, 14(1):5–51.
Wang, K. W., Dumas, M., and Ouyang, C. (2007). Specifi-
cation and analysis of adapters for conversational ser-
vices. eprints.qut.edu.au.
Weber, B., Reichert, M., and Rinderle-Ma, S. (2008).
Change patterns and change support features enhanc-
ing flexibility in process-aware information systems.
Data & knowledge engineering, 66(3):438–466.
Zeng, Q., Sun, S. X., Duan, H., Liu, C., and Wang, H.
(2013). Cross-organizational collaborative workflow
mining from a multi-source log. Decision Support
Systems, 54(3):1280–1301.
Process Adaptation Patterns for Cross-Organizational Business Process Modeling
129