Determining the Progress of a Business Object Based on Its Object

Instances: An Empirical Study

Lisa Arnold

, Marius Breitmayer

and Manfred Reichert

Institute of Databases and Information Systems, Ulm University, James-Franck-Ring, 89081 Ulm, Germany

ﬁ

Keywords:

Object-Centric Business Process, Business Process Monitoring, Business Object, Process Progress.

Abstract:

A fundamental task of any business process monitoring component is to continuously determine the progress

of the running processes of an enterprise. This is particularly challenging when facing dynamic processes

undergoing changes during run-time, which most likely affect the progress of the respective processes as well.

This paper considers object-centric business processes, which consist of business objects and their relations.

During run-time, these business objects may be instantiated multiple times to form object instances. The run-

time behaviour of these object instances is manifested in terms of object lifecycles that interact with each other.

For monitoring a single business object ﬁve alternative methods are introduced, which allow determining the

progress based on average calculations, information about the semantic object relations (hierarchical order,

minimal and maximal cardinality), or event logs (if available). For all methods, the precalculated progress

of individual object instances is leveraged. To evaluate the different methods, an empirical study with 65

participants was conducted. As key observation, the majority of the participants that are experienced with

process modelling and monitoring tools, prefer deriving the progress of a business object from event logs. The

results of this paper are fundamental for determining the progress of a holistic object-centric business process.

1 INTRODUCTION

Monitoring the progress of running processes in real-

time (e.g. in terms of a monitoring dashboard) is

indispensable for agile enterprises. Monitoring al-

lows economising resources (e.g. human and ma-

terial consumption), efﬁciently planning the running

process(es), and saving costs and time. With the in-

sights gained through process monitoring future pro-

cess runs can be optimised. Process progress deter-

mination is a fundamental, but often neglected task

in business process monitoring. In this paper, the

progress determination of a business object in the con-

text of an object-centric business process is deﬁned.

The latter consists of business objects and their re-

lations. During run-time, multiple interacting object

instances are instantiated from these business objects.

Furthermore, the run-time behaviour of these object

instances are deﬁned by their object lifecycles.

Process progress determination is challenging due

to the multitude of object instances that may be cre-

https://orcid.org/0000-0002-2358-2571

https://orcid.org/0000-0003-1572-4573

https://orcid.org/0000-0003-2536-4153

ated (and eventually be deleted afterwards) during

the execution of an object-centric process. Conse-

quently, for a real-world entity, hundreds or thousands

of corresponding object instances and their interac-

tions with instances of other business objects may ex-

ist. In addition to this challenge, dynamic process

changes (e.g. in the context of a process model evolu-

tion) might become necessary affecting the progress

of running object instances as well (Andrews et al.,

2021). Furthermore, at build-time, the total number

of the object instances involved in an object-centric

business process is not fully known, but evolves dur-

ing run-time, varying across the object instance of an

object-centric process. Consequently, the behaviour

of the overall business process (i.e. the object-centric

process) may vary as well (Steinau et al., 2021).

To tackle the challenge of progress determination,

a bottom-up approach is pursued that starts with the

determination of a single object instance and leads to

the determination of the overall progress of a busi-

ness process. This paper deals with determining the

progress of one business object (e.g. object Applica-

tion of a recruitment process) based on the precalcu-

lated progress of its object instances (Arnold et al.,

2021) (e.g. Application1 and Application2) in the

Arnold, L., Breitmayer, M. and Reichert, M.

Determining the Progress of a Business Object Based on its Object Instances: An Empirical Study.

DOI: 10.5220/0012713200003753

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 19th International Conference on Software Technologies (ICSOFT 2024), pages 315-322

ISBN: 978-989-758-706-1; ISSN: 2184-2833

315

context of an object-centric business process (i.e. a

collection of concurrently executed, but independent

object instances). To address this challenge ﬁve alter-

native methods are developed and compared. These

methods are based on the average progress of context-

related object instances, information about semantic

object relations (hierarchical order, minimal and max-

imal cardinality), or event log data (i.e. the process

history). The information (e.g. log data) on which

these methods are based is not always available.

Sec. 2 provides background information about

object-centric processes and sets out the research con-

text of determining the progress of an object-centric

business process. Sec. 3 then introduces ﬁve alter-

native methods for determining the progress of one

business object based on the progress of its context-

related object instances. Sec. 4 describes the research

design of the conducted evaluation study to assess the

ﬁve methods. Sec. 5 analyses and evaluates the re-

sults of this study. Sec. 6 addresses related work and

Sec. 7 concludes the paper.

2 BACKGROUNDS AND

RESEARCH CONTEXT

In the object-centric process management paradigm

a business process is described in terms of interact-

ing business objects that correspond to real-world en-

tities. An implementation of this paradigm is pro-

vided by the PHILharmonicFlows framework, which

enables dynamically evolving object-centric business

processes that allow for both build-in ﬂexibility and

ad-hoc process changes during run-time (Andrews

et al., 2021). The relations between the business

objects, including their cardinalities and hierarchi-

cal structuring, are manifested by the Relational

Process Structure (RPS) (Steinau et al., 2018).

The RPS corresponding to a recruitment business

process is shown in Fig. 1. At run-time, Object

Instances (e.g. Job Offer1 to Job Offer4) of a busi-

ness object (e.g. Job Offer) are created and organised

in a relational instance structure. An example of such

a relational instance structure is shown in Fig. 2.

2.1 Research Context and Research

Questions

Determining the progress of an object-centric busi-

ness process as depicted in Fig. 2 is a challenging

task. First of all, it should be possible to determine the

progress of a single object instance (e.g. Job Offer1)

based on its lifecycle process. This task has already

Figure 1: RPS at design-time with its Business Objects.

Figure 2: Relational instance structure at run-time.

been addressed by us in (Arnold et al., 2021). The

method we presented in this work is based on a one-

dimensional Kalman Filter which is used to determine

the progress of a single object instance. Fig. 2 shows

an example of an object-centric business process em-

phasising the progress of object instances depending

on the progress of their subordinated object instances.

The aim of this paper is to determine the progress of

one business object (e.g. object Review) based on the

precalculated progress (cf. (Arnold et al., 2021)) of its

object instance (e.g. Review1 to Review9). Note that

in object-centric processes, an arbitrary number of ob-

ject instances may be created or deleted during run-

time, which turns the task of determining the progress

of such processes into a challenging endeavour:

• For different business objects there may be a vary-

ing number of object instances and corresponding

object instance interactions.

• The total number of object instances is not com-

pletely known at build-time, but dynamically

evolves during run-time.

• Due to dynamic changes of an object-centric pro-

cess (e.g. to add or delete object attributes or to

change cardinalities of object relations) the be-

haviour of different object instances of an object-

centric process may vary signiﬁcantly.

ICSOFT 2024 - 19th International Conference on Software Technologies

316

To tackle these challenges, the following research

question (RQ) is considered:

RQ: How can the progress of one business object

with multiple object instances be determined in

the context of an object-centric business pro-

cess?

First, different methods for determining the

progress of one business object based on the already

precalculated progress of its object instances are pre-

sented (cf. Sub-RQ 1). Second, the different methods

are investigated and compared in an empirical study

(cf. Sub-RQ 2). The study focuses on the suitability

of the progress determination methods and elaborates

on whether they match human intuition.

Sub-RQ 1: What alternative methods exist to deter-

mine the progress of one business ob-

ject based on the progress of its context-

related object instances?

Sub-RQ 2: Which of these methods is suited best

for users, i.e. matches human intuition

best?

3 PROGRESS DETERMINATION

METHODS

Determining the progress of a varying number of ob-

ject instances created in the context of one business

object can be accomplished in various ways. More

speciﬁcally, progress calculations may be based on

the given RPS (cf. Sec. 2), on average calcula-

tions, or on event log data (i.e. the process his-

tory). This section introduces ﬁve alternative methods

for determining the progress of one business object

(progress

method

) based on the precalculated progress

(cf. (Arnold et al., 2021)) of its object instances.

3.1 Method 1: Total Average

The most intuitive method to determine the progress

of one business object (e.g. Review) is to calculate

the average (AVG) progress of its object instances.

Therefore, the precalculated progress prog

of each

object instance i is added up and the resulting num-

ber is divided by the total number I of these object

instances. This calculation is formalised by Formula

(1).

progress

AVG

∑

i=1

prog

with prog

∈ {x | x ∈ R, 0 ≤ x ≤ 100}

and I ∈ N

(1)

3.2 Methods 2 and 3: Cardinalities

At design-time, each object is related to at least one

other object in the RPS. Corresponding relations are

equipped with a 1:n cardinality (n ∈ N). Thus, for a

particular parent object, many child objects may be

created at run-time. Moreover, the respective cardi-

nality may be restricted by a minimum and maximum.

For example, in Fig. 1 the cardinality between parent

object Application and its child object Review is given

by 1:3..5. This means that, only a maximum number

of ﬁve Review instances may be created in the con-

text of one given Application instance at run-time and

a minimum number of three object instances are re-

quired to terminate the respective business process.

Note that the m:n cardinality (n, m ∈ N) does not exist

in our modelling tool PHILharmonicFlows.

3.2.1 Method 2: Minimum Cardinality

When using the minimal cardinality (MIN), the

progress is determined most optimistically by expect-

ing the minimum number of possible object instances

to be executed, as deﬁned by the minimum cardi-

nality. More speciﬁcally, the precalculated progress

prog

of all object instances i (with the total number of

instances I) are added up and divided by the result of

adding up the maximum max() of the minimum car-

dinality c

min

or the number of created object instances

I(p) for each parent object p (with the total number

of parent objects P). This calculation is formalised by

Formula (2).

progress

MIN

∑

i=1

prog

∑

p=1

max(c

min

, I(p))

with prog

∈ {x | x ∈ R, 0 ≤ x ≤ 100}

and I,P, c

min

, I(p) ∈ N

(2)

3.2.2 Method 3: Maximum Cardinality

When using the maximum cardinality (MAX), the

progress is determined most conservatively by expect-

ing the maximum number of possible object instances

to be executed, as deﬁned by the maximum cardi-

nality. More speciﬁcally, the precalculated progress

prog

of all object instances i (with the total number of

instances I) are added up and divided by the product

of the maximum cardinality c

max

and the total num-

ber of parents P. This calculation is formalised by

Formula (3)

progress

MAX

∑

i=1

prog

max

∗ P

with prog

∈ {x | x ∈ R, 0 ≤ x ≤ 100}

and I,c

max

, P ∈ N

(3)

Determining the Progress of a Business Object Based on its Object Instances: An Empirical Study

317

3.3 Method 4: Event Logs

Event logs (LOG) can be used to predict the expected

number of object instances E that will be created for a

business object in the context of their common parent

object instance. For this purpose, all ﬁnished parent

object instances (of either terminated or running pro-

cesses) are continuously analysed and the expected

number of created object instances is determined (e.g.

based on average or machine learning). To calculate

the progress based on this estimation, the precalcu-

lated progress prog

of all object instances i (with the

total number of instances I) are added up and divided

by the result of adding up the maximum max() of the

expected number of object instances E or the number

of created object instances I(p) for each parent object

p (with the total number of parent objects P). Note

that this method may only be used if an event log ex-

ists and the accuracy of the prediction depends on the

quality of the log. This calculation is formalised by

Formula (4).

progress

LOG

∑

i=1

prog

∑

p=1

max(E, I(p))

with prog

∈ {x | x ∈ R, 0 ≤ x ≤ 100}

and E ∈ R and I, P, I(p) ∈ N

(4)

3.4 Method 5: Parent Object

The Parent Object method (PO) combines two con-

cepts:

1. Considering the object instances created for a

business object in the context of each parent ob-

ject instance individually (as in MIN, MAX, and

LOG).

2. The average calculation (as in AVG).

For this, ﬁrst the average of the progress prog

i∈p

of all object instances I(p) created for a business ob-

ject in the context of each parent object instance p

(e.g. all reviews of the same application) is calcu-

lated individually. Second, the average of these re-

sults is calculated (with the total number of parent ob-

ject instances P) to determine the overall progress of

the business object (e.g. object Review). Note that

prog

i∈p

numbers the progress of the instances i for

each parent object instance individually. This calcu-

lation is formalised by Formula (5).

progress

∑

p=1

∑

I(p)

i=1

prog

i∈p

I(p)

with prog

i∈p

∈ {x | x ∈ R, 0 ≤ x ≤ 100}

and I(p), P ∈ N

(5)

4 RESEARCH METHOD

This section summarises the research method under-

lying the empirical study we conducted to assess the

ﬁve progress determination methods from a human

perspective. In detail, this section focuses on the data

collection method, the study design and structure, and

the data analysis method.

4.1 Data Collection

How users perceive the progress of a business ob-

ject based on the ﬁve progress determination meth-

ods is investigated in an empirical study. For conduct-

ing this study and collecting data, the web-based tool

Unipark is leveraged. The study is performed based

on an anonymous online questionnaire

and is avail-

able in both English and German language. More-

over, both language options do not differ with respect

to content or structure. The questionnaire was avail-

able over a period of one month.

4.2 Study Execution

The empirical study is structured in ﬁve parts with a

total of 35 questions. For this purpose, dichotomous,

semi-open multiple choice questions, and open ques-

tions are used. Moreover, some of the questions are

identical, but refer to different backgrounds.

Demographics and Experience. In the ﬁrst

part of the study, demographic data is queried from all

participants. This includes information like gender,

age, current profession, and professional ﬁeld. More-

over, the experiences of the participants in respect to

process modelling and (business process) in general

are queried.

Training. The second part offers a training ses-

sion for about half of the participants. The other half

does not participants in any training session. For se-

lecting the participants a random function is used.

The training includes the explanation of the progress

determination methods as well as the study structure.

Perception of Progress. The main part of

the study aims to assess the perception of progress for

three different scenarios using the described progress

methods (cf. Sec. 3). For each scenario multiple

questions with a varying number of object instances

are given.

Questionnair and responses of the 65 participants are

available on Researchgate: https://www.researchgate.net/

publication/378140057 Determining the Progress

ICSOFT 2024 - 19th International Conference on Software Technologies

318

Details about the scenarios.

1.) Evaluation of research papers for a conference.

All papers shall be evaluated by 3 to 5 reviewers to

decide whether or not the paper can be accepted.

2.) Grade bonus for students. To receive the grade

bonus for a particular exam, students need to

achieve at least 80% of all exercise points from

the 12 to 15 exercise sheets.

3.) Recruitment process. A company has published a

large number of vacancies. Depending on the job

offer, there are many, few, or no applicants.

With 14 semi-open multiple choice questions (i.e.

seven for Scenario 1, three for Scenario 2, and four for

Scenario 3), the participants have to choose their most

appropriate progress calculated for each question of

the given scenario. Additionally, for each progress,

the calculated progress determination method is pro-

vided (i.e. ”a.) 85% (AVG)”). Moreover, if none of

the given methods match the participants intuition of

progress, the participants may indicate their percep-

tion of progress with a short explanation.

Cognitive Strain and Behaviour during

Participation. Each scenario is completed with

the following three questions to investigate the

behaviour of the study participants.

1.) Were you able to answer the previous questions

clearly? If not, why?

2.) Did you change answers from previous questions

when answering this form sheet? If yes, why and

where?

3.) Have you chosen different methods for different

scenarios? If yes, why?

Ranking the Methods: Additional to the

scenarios, the participants rated the progress deter-

mination methods in a ”drag-and-drop” like manner,

with ranking the method they consider being most

suitable at the top. As not all methods are always

applicable to determine progress, the following

three combinations are considered to cover the most

common real-world conditions.

Available methods:

Case 1: AVG, MIN, MAX, LOG, and PO

Case 2: AVG, MIN, MAX, and PO

Case 3: AVG, MIN, and LOG

First, all methods are available for the participants to

deﬁne a suitable ranking. Second, all methods except

LOG is given to ﬁnd a suitable ranking if no event

log exists. Third, MAX and PO are not available, as

the maximum cardinality is often not deﬁned for top-

level objects and the PO is not applicable to top-level

objects in general. Additionally, an optional text ﬁeld

is given to describe a method or procedure that does

not corresponds to the ﬁve introduced progress deter-

mination methods.

Language: Finally, the participants are asked,

which language version they read to verify that no dif-

ferences exist in the translation and wording.

4.3 Data Analysis

The study structure and its data analysis and valida-

tion are generated on the checklist of the empirical

cycle described in (Wieringa, 2014). Furthermore,

all collected data of the questionnaire are analysed

and evaluated based on the methodology presented

in (Wieringa, 2014; Brace, 2018). The aim of the

evaluation is to ﬁnd a representative answer that re-

ﬂects the opinion of the participants. In detail, an ex-

ploratory analysis is applied that uses Cross Tabula-

tions to compare the quantitative results from differ-

ent participant groups. In addition, open-ended ques-

tions are extended to investigate the cognitive strain

and behaviour during participation. In the evalua-

tion, different participant groups are deﬁned accord-

ing to their background and prior knowledge. The di-

vision into groups allows to investigate variations in

the perception of progress and the associated choice

of progress determination methods depending on the

background knowledge of the individual participants.

5 EVALUATION

Overall, 65 participants completed the questionnaire

of the empirical study. 5 of them were excluded as

they had problems understanding the tasks (answer-

ing the questions as ”not clear” at cognitive strain

part). The following evaluation is based on the an-

swers of the remaining 60 participants.

5.1 Demographics and Experience

In total, more male (39 | 65%) than female (19 |

31.7%) participants took part. The remaining partici-

pants deﬁne their gender as non-binary (2 | 3.3%). Al-

together, the participants are between 19 and 34 and

on average 25.13 years old.

Most participants have their profession ﬁeld in the

MINT (Mathematics - Computer Science - Natural

Science - Technology) (51 | 85%). The remaining par-

ticipants have a background in economics (9 | 15%).

Thereby, most participants are students (43| 71.7%)

studying either in a MINT (34 | 79,1%) or an eco-

nomics program (9 | 20.9%). The second major share

Determining the Progress of a Business Object Based on its Object Instances: An Empirical Study

319

is given by academics (14 | 23,3%) in MINT. The re-

maining participants are working in industry with fo-

cus in MINT (3 | 5%). With this group of participants,

confounding variables (as e.g. general school educa-

tion or major age differences) could be avoided as far

as possible.

Exactly one quarter (15 | 25%) of the participants

have no experiences with process modelling, most of

them are students (14 | 93.3%). Half of the partic-

ipants (30 | 50%) are not familiar with monitoring

tools, 27 of them are students (27 | 90%). More than

one third (21 | 35%) of the participants have experi-

ences with business process monitoring tools. This

group is also familiar with process modelling and

monitoring tools and deﬁne the expert group. In this

expert group, the share of male (10 | 47,6%) and fe-

male (11 | 52.4%) is similar. On average they are 26.8

years. Most of them have their professional ﬁeld in

MINT (18 | 85.7%) the remaining in economics (3 |

14.3%). Most of them are academics (11 | 52.4%).

The remaining participants are divided in students (9

| 42.8%) and others (1 | 4.8%). The non-experts are

composed of all participants except the experts. In the

following, the results of the experts (E) are compared

with the ones of the non-experts (NE) to enable a pro-

found analysis.

5.2 Perception of Progress

The perception of progress differs signiﬁcantly be-

tween the experts and non-experts (cf. Tab. 1). The

biggest difference can be found in the number of par-

ticipants that chose the progress calculation based on

LOG. In each of the 14 perception progress questions,

at least 71.4% (up to 85.7%) of experts chose the

progress based on LOG. In contrast, the non-expert

selected this method only with at least 23.1% (up to

38.5% in the second scenario and up to 51.3% in the

third scenario). In 5 of 13 cases, where AVG and LOG

are used, the non-experts opted for AVG. In the other

8 cases, the majority favoured LOG. Consequently,

LOG has been mostly chosen, when object instances

are created from more than one parent object.

In the ﬁrst scenario, the answers of the non-

experts are approximately evenly distributed between

the progress determination methods. The methods

AVG, MIN, and PO are chosen less and less over the

questionnaire (Scenarios 2 and 3). Considering the

non-experts for the ﬁrst scenario at most 10.3% pre-

ferred the progress based on MAX. However, in the

second scenario this method was selected by up to

28.2% of the non-experts. This may be due to the

fact that in the ﬁrst question of Scenarios 2 only 3 of

at least 12 (up to 15) exercise sheets have been com-

pleted and the progress is calculated far too high when

choosing AVG. Accordingly, AVG was chosen signif-

icantly (above 10%) less. As another observation that

emerged from the study responses, PO is not preferred

by experts. In comparison, up to 20.5% of the non-

experts chose PO. Furthermore, the number of partic-

ipants choosing the same method for all questions of

one scenario is increasing signiﬁcantly from the ﬁrst

to the third scenario (to threefold) by the non-experts.

In general, it can be observed that non-experts do not

have a clear preferred method. In contrast, LOG was

preferred by the experts in most cases (more than 80%

in Scenarios 1 and 2, more than 70% in Scenario 3).

The last question of scenario 3 addresses the issue of

how to deﬁne progress if no object instance is created.

In this case, both groups answer very similar. Above

70% voted for 0% and the remaining for 100%. Addi-

tionally, one expert remarked that no progress should

be assigned in this case and the progress of the parent

object should be used instead.

5.3 Ranking the Methods

In this part, the participants rank the progress determi-

nation methods considering their suitability. There-

fore, the position of each method is represented by a

number. For example, 1 is assigned to the top-rated

method.Tab. 2 shown, the average position for each

method for both groups separately.

PO is rejected by most participants in both groups.

For example, non-experts rated it with 4.2 out of 5

(whereas 5 is most unsuitable) and the experts rated it

with 4.6 out of 5 in the ﬁrst case. The remaining rank-

ing has shown similar results. Note that the number

of available methods differs in Case 2 and 3. Further-

more, MIN is evaluated very similarly by both groups

and is rejected, next to PO, as the second most unsuit-

able method. Compared to MIN, MAX is more suit-

able. In general, most participants preferred a more

conservative progress calculation.

Both groups differ in their choice of the most suit-

able method. The non-experts selected AVG (1.6 to

2.1) as their favoured method in each case and the ex-

perts chose LOG (1.3). Note that AVG only consid-

ers existing object instances (snapshot) whereas LOG

also considers object instances that are expected to be

created in the future (big picture). This supports the

statement that the experts have the big picture of a

business process in mind and not only consider the

snapshot of the current process. In contrast, the non-

experts are more focused on the snapshot as on the

big picture. Due to this fact, the experts evaluated

MAX (1.9) as the best suitable option, that should be

used when LOG is not available. The non-expert’s

ICSOFT 2024 - 19th International Conference on Software Technologies

320

Table 1: Study results in percent for each question of the three scenarios. Table entries marked with ”-” were not part of the

response options in the study.

Scenario 1 Scenario 2 Scenario 3

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q1 Q2 Q3 Q1 Q2 Q3

AVG NE 33.3 48.7 35.9 20.5 28.2 20.5 35.9 20.5 25.6 17.9 56.4 48.7 46.2

E 9.5 4.8 9.5 4.8 4.8 4.8 4.8 4.8 9.5 9.5 23.8 28.6 28.6

MIN NE 30.8 17.9 23.1 28.2 15.4 17.9 7.7 12.8 17.9 20.5 - - -

E 4.8 9.5 0 4.8 4.8 4.8 4.8 4.8 4.8 0

MAX NE 7.7 10.3 10.3 10.3 7.7 7.7 7.7 28.2 25.6 20.5 - - -

E 4.8 4.8 9.5 0 0 0 0 9.5 4.8 4.8 - - -

LOG NE 28.2 23.1 30.8 25.6 33.3 33.3 33.3 38.5 30.8 33.3 41 51.3 48.7

E 81 81 81 85.7 81 85.7 85.7 81 81 85.7 71.4 71.4 71.4

PO NE - - - 15.4 15.4 20.5 12.8 - - 5.1 - - 5.1

E - - - 0 0 0 0 - - 0 - - 0

Table 2: Average position in the ranking for the three given cases. Results are coloured to visualise the better or worse rating

comparing both groups.

Case 1 Case 2 Case 3

AVG MIN MAX LOG PO AVG MIN MAX PO AVG MIN LOG

NE 2.1 3.4 3.0 2.2 4.2 1.8 2.8 2.2 3.3 1.6 2.6 1.8

E 3.0 3.5 2.7 1.3 4.6 2.1 2.5 1.9 3.5 2.2 2.5 1.3

choice of the most appropriate method (AVG) is al-

ways available. Considering the distribution of the av-

erage position, the range of the experts varies largely

(e.g. 1.3 to 4.6 for Case 1). Consequently, most par-

ticipants rank the methods in the same way. In con-

trast, the range of the non-experts is smaller (e.g. 2.1

to 4.2 for Case 1). This indicates that the selected

ranking of the methods differs among the non-experts,

which leads to similar results for different methods.

For example, in Case 1, AVG results in 2.1 and LOG

in 2.2 on average. In the following, the total ranking

for both experts and non-experts are given. Note that

this ranking is the same for all three cases indepen-

dent of non-available methods.

Non-expert: AV G > LOG > MAX > MIN > PO

Expert: LOG > MAX > AVG > MIN > PO

5.4 Training and Language

In total, 45 (69.2%) of the participants took part in

a previous training. In general, no differences could

be found between participants with and without train-

ing. Furthermore, only 14 (21.5%) participants read

the English questionnaire whereas about one quarter

are experts (4 | 28.6%). The distribution of experts

and non-experts regarding the language choice is al-

most identical to their distribution and no differences

could be found in the answers.

5.5 Limitation of the Study

In the ﬁrst scenario, participants often chose different

methods for the individual questions. However, as the

study progressed, participants increasingly chose only

one or two methods. As the order of the scenarios was

the same for each participant it is unclear whether this

observation is based on a learning progress of the par-

ticipants or on the scenarios themselves. Furthermore,

the scenarios where rather simple to allow for a bet-

ter understanding. However, the transferability of the

results to complex scenarios was not investigated.

6 RELATED WORK

The current research in the ﬁeld of OCEL (object-

centric event log) (Ghahfarokhi et al., 2021) allows an

event log to be related to multiple objects that means

each row is given by one object and includes, for ex-

ample, its identiﬁer and type of object. Traditionally,

an event log is event-based that means each row in a

table has at least an identiﬁer, an activity, a timestamp,

and related objects. The OCEL is similar to the event

log generated from our PHILharmonicFlows frame-

work. Our approach can be applied to OCEL if the

progress of the individual instances is given. This pre-

condition also exists for object-centric business pro-

Determining the Progress of a Business Object Based on its Object Instances: An Empirical Study

321

cesses, but was already addressed in (Arnold et al.,

2021). In (Gherissi et al., 2022), an approach for pre-

dictive process monitoring based on OCELs is dis-

cussed. In general, this approach improves the ac-

curacy in predicting the next activity and the MAE

(Mean Absolute Error) in time prediction compared

to the conventional event logs by utilising the interac-

tion between objects.

7 SUMMARY AND OUTLOOK

In this paper, two research questions were addressed

to determine the progress of a collection of object in-

stances created in the context of one business object.

Regarding Sub-RQ 1, ﬁve possible progress determi-

nation methods were introduced and formally deﬁned.

To address Sub-RQ 2, an empirical study was con-

ducted that investigates the most suitable progress de-

termination method. In this context, progress deter-

mination was considered for three different scenarios.

In addition, a method ranking was created. This rank-

ing helps to decide which progress method will be ex-

ecuted ﬁrst, and if this method is not an option due to

non-existing conditions (e.g. the event log for LOG)

which is the next favoured. For the evaluation, an ex-

pert group, which comprises the participants with ex-

periences in process modelling and business process

modelling tools, and a non-expert group, consisting

of the remaining participants, are deﬁned. In general,

non-experts have no clearly preferred method as they

prefer both AVG and LOG. The experts agree in all

cases (with more than 70%) and consider LOG to be

the most appropriate. Considering this result, LOG

should be used to determine the progress method.

When the information required by this method is not

available progress determination should be based on

MAX. Finally, if the maximum cardinality is not im-

plemented, AVG should be used.

Future research is needed to provide fully func-

tional progress calculation in a monitoring tool for

object-centric business processes. In addition, the

usability and accuracy of their resulting outcomes

will be tested directly at the monitoring tool by end-

users. For example, a Delphie study with focus on

experts more complex scenarios might provide addi-

tional insight into the usability and human intuition of

progress determination. The run-time behaviour (e.g.

response times) of the individual methods is exam-

ined in order to check their suitability directly in our

monitoring tool. The combination of methods is also

being investigated in order to achieve more precise

progress.

ACKNOWLEDGEMENTS

This work is part of the ProcMape project, funded

by the KMU Innovativ Program of the Federal Min-

istry of Education and Research, Germany (F.No.

01IS23045B).

REFERENCES

Andrews, K., Steinau, S., and Reichert, M. (2021). En-

abling runtime ﬂexibility in data-centric and data-

driven process execution engines. Information Sys-

tems.

Arnold, L., Breitmayer, M., and Reichert, M. (2021). A

one-dimensional kalman ﬁlter for real-time progress

prediction in object lifecycle processes. In 2021

IEEE 25th International Enterprise Distributed Ob-

ject Computing Workshop (EDOCW), pages 176–185.

IEEE.

Brace, I. (2018). Questionnaire design: How to plan, struc-

ture and write survey material for effective market re-

search. Kogan Page Publishers.

Ghahfarokhi, A. F., Park, G., Berti, A., and van der Aalst,

W. M. (2021). Ocel: A standard for object-centric

event logs. In European Conference on Advances in

Databases and Information Systems, pages 169–175.

Springer.

Gherissi, W., El Haddad, J., and Grigori, D. (2022). Object-

centric predictive process monitoring. In International

Conference on Service-Oriented Computing, pages

27–39. Springer.

Steinau, S., Andrews, K., and Reichert, M. (2018). The

relational process structure. In Int. Conf. on Ad-

vanced Information Systems Engineering, pages 53–

67. Springer.

Steinau, S., Andrews, K., and Reichert, M. (2021). Coor-

dinating large distributed relational process structures.

Software and Systems Modeling, 20(5):1403–1435.

Wieringa, R. (2014). Design science methodology for infor-

mation systems and software engineering. Springer.

ICSOFT 2024 - 19th International Conference on Software Technologies

322