A Process Mining Methodology for People Analytics:

With a Case Study on Recruitment Analysis

Emiel Caron and Nijs Niessen

Tilburg University, Department of Management, Warandelaan 2, PO Box 90153, Tilburg, The Netherlands

Keywords: HR Analytics, People Analytics, People Management, Process Mining, Process Mining Methodology.

Abstract: In today’s competitive business environment, acquiring data about Human Resource (HR) processes and

optimizing operational excellence are some of the main objectives. Process mining as a specific form of HR

Analytics aims at translating data, stored in the HR information systems of the organizations into insights

about the organization’s HR processes. These insights are then translated into improvements for HR processes

and to enhancements in the compliance to rules and regulations. However, so far there is no concrete and

comprehensive framework for the execution of process mining analysis that can be used practically by the

HR professionals. In this paper, we develop a methodology for HR Analytics, based on PM

; a step-by-step

method with the best practices and approaches to process mining projects in the HR domain. Additionally,

we establish definitions such as the HR event log, the business process, etc. that capture the specifics of the

HR domain. Finally, we demonstrate the effectiveness of the proposed methodology by applying it to a case

study on the recruitment process. The results show that the methodology successfully identifies areas for

improvement and provides insights that can enhance the overall HR recruitment process.

1 INTRODUCTION

Human Resource Management (HRM) as a business

function aims to effectively deploy and manage

employees within an organization in order to gain a

competitive advantage. A new trend within the field

of Human Resources (HR) is the use of HR Analytics

(HRA) or people analytics (Marler and Boudreau,

2017), which is, as opposed to the traditional HRM, a

data-driven approach to better understand and

optimize HR processes and practices, like recruitment

and selection, performance management, employee

turnover, training and development, and so on.

HR analytics are founded on the Human Resource

Information Systems (HRIS). Chauhan, Sharma &

Tyagi (2011) define the HRIS as: “[…] the

integration of software, hardware, support functions

and system policies and procedures into an automated

process designed to support the strategic and

operational activities of the human resources

department and managers throughout the

organization.” By providing timely, accurate and

understandable information the HRIS’s support

decision-making for HR professionals. HRIS serve as

integral components of the business processes they

support. Business processes are defined as a chain of

linked activities that transform inputs into value

added output to achieve the company’s objectives, by

applying one or more resources (Christensen et al.,

2016).

Potential techniques to produce HR analytics, that

can utilize the process data within HRIS, are process

mining techniques (Aalst, 2012), that focus on the

discovery, monitoring and improvement of processes

by extracting knowledge from event logs in

information systems. In this research, we want to

engage with three process mining research

challenges, identified by R'bigui and Cho (2017), by

using processing mining in an HR context. These are:

1) combining process mining with other type of

analysis and to enrich event log data, 2) improving the

usability for non-experts, in this case HR

professionals, and 3) improving the understandability

for non-experts, to produce process models that can

be used for HR decision-making.

Moreover, the objective of this research is

fourfold: 1) to extend and specialize PM

- the process

mining project methodology (Van Eck et al., 2015) -

for the specific development of HR and people

analytics, 2) to define the concept of an event log in

the case of a HR information system, 3) to apply the

specialized methodology on a real case study on

Caron, E. and Niessen, N.

A Process Mining Methodology for People Analytics: With a Case Study on Recruitment Analysis.

DOI: 10.5220/0012135100003538

In Proceedings of the 18th International Conference on Software Technologies (ICSOFT 2023), pages 645-651

ISBN: 978-989-758-665-1; ISSN: 2184-2833

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

645

Figure 1: PM

2HR

, Process mining methodology for HR, adapted from Van Eck et al. 2015.

employee recruitment and selection, and 4) to

enhance the current limited scientific literature (Arias

et al. 2018, Jooseok et al. 2018) on the utilization of

process mining in the field of HRM.

The remainder of this paper is structured as

follows. In Section 2 we develop the process mining

framework for people analytics. Subsequently, in

Section 3 we apply the framework on a case study for

recruitment analysis. Conclusions are drawn in

Section 4.

2 METHODOLOGY FOR HR

ANALYTICS: PM

2HR

We propose a specialisation of the six stages (Section

2.1-2.6) of the PM

process mining project

methodology (Van Eck et al., 2015), to make it useful

for HR processes and analytics, called PM

2HR

. The

initial planning of the project consists of: the

definition of the HR decision problem and the related

data extraction from the HR data source (stages 1-2).

After that extracted event data then goes through

several analysis iterations. In each iteration, HR data

processing, process mining & analysis, and

evaluation are executed (stages 3-5). Lastly, gained

insights are used to modify the actual HR business

process (stage 6) that takes place in the company, i.e.

process improvements as part of business process

management. An overview of PM

2HR

is shown in

Figure 1. In the following sub sections we discuss HR

specific perspectives for the six stages.

2.1 Stage 1. Define the HR Business

Problem

The goal of the planning step is to start the people

analytics project and determine the initial HR

business questions. Van Eck et al. (2015) define three

activities in this stage: identifying business questions,

selecting business processes, and composing a project

team. People analytics, can help answer a variety of

HR-related business questions. Example HR business

process and related business questions are:

 Recruitment and selection: What is the time-to-

hire for open positions? How many candidates

does it take to make a successful hire?

 Employee onboarding: How long does it take

for new employees to become fully productive?

Which onboarding activities are most effective

in achieving this?

 Performance management: How often are

performance reviews conducted? What is the

average performance rating across the

organization?

 And so on for other HR business processes that

are supported HRIS like training and

development, compensation and benefits,

employee engagement, turnover and retention.

Moreover, the PM

2HR

methodology addresses two

important research challenges (R'bigui and Cho,

2017): enhancing the usability and understandability

of process mining models for non-experts,

specifically HR practitioners. By involving HR

practitioners directly in the project team alongside

process mining experts, both the managerial and

analytical perspectives are integrated from the outset.

This approach results in process models that better

align with HR objectives.

2.2 Stage 2. Extract HR Data

In the data extraction stage the relevant event data,

and optionally ex-ante process models, are extracted

from the HRIS, that support the HR processes

identified in the previous stage. This stage consists of

three activities (Van Eck et al., 2015): determining

scope, extracting raw event data and transferring

process knowledge. HRIS provide a wide variety of

functions in a company, Chauhan et al. (2011)

separate these functions into three categories:

ICSOFT 2023 - 18th International Conference on Software Technologies

646

operational, tactical, and strategic. HRIS are often,

partially, process-aware information systems (Van

der Aalst, 2009), as they are driven on the basis of HR

process models (see Stage 1). In this step the output

is the HRIS database, that captures all raw data

required to answer the business question, so specific

event log data and other relevant HR data.

2.3 Stage 3. Processing HR Data

After the initialization phase, the analysis iterations

begin with the processing of the data from the HRIS

database. The objective of this stage is to create HR

event logs from the extracted HRIS data and prepare

them for the mining & analysis stage. Processing

activities include (Van Eck et al., 2015): creating

views, aggregating events, enriching logs and

filtering logs.

Here we focus on the generic data requirements

for using process mining in the domain of HRA. The

HR event log is a necessary element for process

mining in the HR domain. We define the HR event log

as: a (processed) data set that satisfies the three

minimum (1-3) properties for process mining as

specified in the XES standard

: 1. a Case ID, 2. an

Activity name, and 3. a Time-stamp, and has

additional properties (4-6) that are related to the HR

decision problem. The additional properties are:

4. Cases relate to people, typically employees. In

HR, process instances are typically related to

employees, where the instances describe the

employee’s life cycle with regards to its recruitment,

internal mobility, educational path, etc.

5. People data needs to follow strict legislation.

This property follows from 4. Since the data is

extracted from the HRIS and it contains data about

people, it is necessary that it complies to certain rules

and regulations, like the General Data Protection

Regulation

(GDPR) in Europe. In particular, some of

its principles include that the data is required to be

processed lawfully and in a transparent manner as

well as accurate and collected for explicit and

legitimate purposes.

6. The inclusion of attributes related to the HR

The HR event log must include attributes that are

specific to the HR process is analyzed. Examples of

HR specific attributes include details related to the:

employee, job, organization, HR process, etc. For

example, the division (i.e. Finance, Marketing) that is

responsible for performing an activity or the job

position of the employee (i.e. Manager, Senior

Manager, Junior Analyst, etc.).

https://xes-standard.org/

2.4 Stage 4. Process Mining & Analysis

In this stage the HR event logs are analyzed using

process mining techniques to answer the research

questions and gain insights into the HR business via

dedicated analytics. In this stage four types of

activities are identified (Van Eck et al., 2015):

process discovery, conformance checking,

enhancement, and process analytics. The first three

activities are the main classes of process mining

techniques (Aalst, 2012). Process analytics constitute

the class of all other analysis techniques, e.g. data

mining, statistics, dashboards and scorecards etc.,

which can be applied in the context of the business

process, often in combination with the first three

activities. In addition, process mining covers different

analytical perspectives (Aalst, 2016): the control-

flow perspective that focuses on the ordering of

activities, the organizational perspective that focuses

on the resource, and the case perspective that includes

all properties of the cases.

From an HR professional's perspective, the

primary goal of process analytics is to develop

specific metrics and key performance indicators

(KPIs) that can be used to manage HR business

processes effectively. These metrics are then

combined into an HR performance scorecard or

dashboard, making them easily accessible and visible

to HR professionals. By doing so, the focus shifts to

developing metrics that directly relate to the business

problem at hand. To compute these metrics, various

classes of process mining techniques are utilized.

In Table 1, we present an example HR

performance scorecard with relevant metrics for the

recruitment process. Muensterman et al. (2009) and

Laumer et al. (2014) define metrics and targets related

to the recruitment process in organizations over the

dimensions: process time (e.g. time-to-hire), process

costs (cost-per-hire) and recruitment process quality.

The HR performance scorecard uses information

from the derived process model, visualized by

process mining and captured in process statistics.

Where some process models can be very complex and

ineligible, the scorecard allows for a structured way

of describing what is happening in different models at

a glance. Similar scorecards can be produced for other

HR business processes.

https://gdpr-info.eu/

A Process Mining Methodology for People Analytics: With a Case Study on Recruitment Analysis

647

Table 1: Example HR performance scorecard with metrics

related to recruitment.

2.5 Stage 5. Evaluation

During this stage, the findings are evaluated to refine

research questions and generate improvement ideas.

This stage includes three main activities: diagnose,

verify, and validate. Diagnosing the findings involves

separating expected and unexpected findings and

refining new research questions. For instance, if the

average time to promotion between two functions is

much shorter than expected, further questions may be

raised to analyse this finding in more detail and

determine whether the effect is global or only

observed in a subset of employees. The next step is to

verify and validate the results. Verification entails

comparing the findings to the original data to ensure

their correctness. For example, verification might

reveal that an unexpected finding was caused by

errors in data preparation. On the other hand,

validation involves comparing the findings to the

claims, hypotheses, or knowledge of stakeholders. In

this activity, HR professionals are crucial to validate

the findings, as process analysts may misinterpret the

results. Validation may also reveal the cause of

unexpected findings, such as when it is discovered

that two functions are very similar and should be

combined into one, as seen in the example above. At

this stage, there are two possible follow-ups: refined

research questions can be investigated in a subsequent

analysis iteration, or improvement ideas can be

applied in the HR decision-making support phase.

Overall, the evaluation stage is essential to ensure that

the process mining findings are valid and reliable, and

that they provide actionable insights for HR process

improvement.

2.6 Stage 6. HR Process Improvement

The final stage relates to use the gained insights to

improve the HR processes. This stage involves two

activities, adapted from van Eck et al. (2015): HR

policy improvement and supporting operations. For

example, if the analysis identifies issues with the

internal mobility of employees, improvement ideas

can be used to guide HR decision-making and modify

the internal mobility process through HR policy

change. It is important to note that improving HR

policy is a different area of expertise from process

mining, and it may require a separate project. One

potential improvement is to change recruitment

policy for certain positions in response to the finding

that external hires often leave while internal hires do

not. This can help to reduce employee turnover and

ensure that the organization retains valuable talent.

The supporting operations activity involves using the

improvement ideas to support daily HR operations.

This can include implementing new tools or processes

that align with the improved HR policies. For

example, if the HR policy is changed to encourage

internal mobility, new training programs can be

developed to help employees adopt new skills and

prepare for new roles.

3 CASE STUDY: INSIGHTS IN

RECRUITMENT PROCESSES

In this section we apply and evaluate PM

2HR

in a case

study to provide insights on the recruitment process

of a company in the global technology industry.

Stage 1. Define the HR Business Problem.

The company's primary focus is to enhance

operational excellence in its recruitment and selection

process, aiming to execute it more strategically,

consistently, and reliably. In simpler terms, the

business goal is to optimize the recruitment process

to achieve better outcomes. The HR professional

defines the business problem. In this case it is about

“gaining insights in the performance of its

recruitment process such that suggestions for process

improvements are developed”, and questions like

“How does the actual recruitment and selection

process look like?”, “Are there any bottlenecks in the

recruitment process?”, “What metrics can be used to

measure the performance of the recruitment process”,

ICSOFT 2023 - 18th International Conference on Software Technologies

648

and “What are suggestions for process

improvements?”.

Stage 2. Extract Recruitment Data.

The data is extracted from the Applicant Tracking

System (ATS) database by SQL scripts. The ATS is

integrated in the HRIS and it represents a specialized

information system for the recruitment process. When

the ATS data is extracted from the HRIS database, the

data is transformed and stored into three different

tables: Requisitions, Applications, and Application

status. The tables contain information about the

requisition ID, the activity performed (i.e. screen

Curriculum Vitae (CV) by recruiter, interview, hire),

which division is responsible for the requisition,

personal details of the applicant (name, date of birth,

education etc.), submission details of the application

and several time-stamps indicating the start and the

end date of the requisition. In this case, the data

extracted from the ATS database, after preparation,

satisfies the properties of the HR event log (see

Section 2.3). Therefore, process mining can be

applied meaningfully.

Stage 3. Prepare Recruitment Event Log.

The extracted data has a csv file format. In order to

use process mining, it is necessary that the data is

transformed to an HR event log. The data

transformation is implemented using Python. The

data structure was altered such that the three different

tables were merged into one single table, to satisfy the

properties of the HR event log. Therefore, the result

of the data processing and transformation is the HR

event log. In particular, it contains the three

requirements of an event log: a Case ID (req.user.id),

an activity name (applicant.status), a timestamp

(start.time, end.time) and other additional attributes

that are listed as other columns: background,

education, division, sub.division, requisition.region.

The HR event log contains data of the years 2018-

2019, with 250,000 cases (applications), nearly

600,000 events (application activity), and 2,500

different process variants.

Stage 4. Process Mining And Analysis.

The next step is to construct a process mining model

using the HR event log. Disco software

is used to

further explore the HR event log, examine its quality

and analyse the recruitment process. In this case, the

HR event log analysis is both explorative and goal

driven. Thus, the HR professional implements a

control flow analysis as well a performance analysis.

First, the normative recruitment process (NRP) is

discovered using fuzzy miner together with the

https://fluxicon.com/disco/

complete recruitment process and secondly the

process efficiency is measured using certain metrics

and ratios. The tasks in the NRP are: Screen CV by

Recruiter → Screen CV by Hiring Manager (HM) →

Interview candidate → Make offer → Hired / Close

application.

Discovering The Actual Recruitment Process.

In Figure 2, a process model for the full recruitment

process is discovered, by using the filtering

parameters to obtain a clear and eligible model. The

main activities are clearly visible and an almost

desirable recruitment is shown.

Figure 2: Discovered (filtered) recruitment process.

A new addition to the activities of the NRP is the

‘Screening interview by Recruiter’ step. In the figure,

some unexpected paths are still visible: Cases going

from ‘Screen CV by Recruiter’ to either ‘Interview’

or ‘Offer’ and cases going from ‘Closed’ or ‘Hired’

back to ‘Screen CV by Recruiter’. When drilling

down on these cases that go from ‘Closed’ to ‘Screen

CV by Recruiter’, the variants show that most of these

steps occur because the case got either re-opened by

the recruiter, or maybe the applicant applied again,

triggering the ATS to change the status to ‘Screen CV

by Recruiter’. In Table 2, the five most frequent

variants for hires are given. Comparing the results to

the variants of the NRP shows that most of the

variants and their frequency stay the same, except that

a new activity is introduced for the desirable variant:

‘Interview 2’ is added.

Process Efficiency: Recruitment Performance.

Following the recruitment process performance

scorecard (see Table 1), we produce the metrics using

the software. Results given in the scorecard, depicted

in Table 3, show that the recruitment process does not

reach the targets set by business experts. The

differences between the target and the actual data

A Process Mining Methodology for People Analytics: With a Case Study on Recruitment Analysis

649

Table 2: Recruitment process hire variants.

Hire variants Relative

Screen CV by Recruite

Offe

Hired 21.65%

Offe

Hired 14.45%

Screen CV by Recruite

Hired 13.58%

Screen CV by RecruiterScreen CV by

HMInterviewInterview2Offe

Hired

9.63%

Screen CV by RecruiterScreen CV by

HMOffe

Hired

2.70%

Other hire variants 23.15%

ranges from 0.5 days to almost 26 days. Especially

the metrics for ‘Time to Select’ and ‘Time Interview

to Offer’ are much higher than their targets, while the

‘Time to Hire’ is only marginally higher than the

target. Following the pipeline requisitions, promising

candidates go through the real requisitions relatively

fast, with a minimum of two days. After further

investigation, applications to these real requisitions

skip activities like ‘Screen CV by HM’ or ‘Interview’,

which are recorded in the pipeline requisition,

resulting in a lower mean score for ‘Time to Hire’ and

exclusion from the ‘Time to Select’ and ‘Time

Interview to Offer’ metric, which are measured from

‘Screen CV by HM’ and ‘Interview’.

Table 3: Recruitment performance scorecard.

Process Efficiency: Educational Perspective.

For an analysis of the educational perspective, a

comparison between the educational levels of the job

functions is made. The three levels are Bachelor,

Master, PhD. When comparing the metrics in Table

4, the scorecard shows that the Bachelor-level

applications are screened faster on average than their

PhD-counterpart and score below average. This could

be the result of the requirements for a Bachelor-level

function are defined more clearly and are more

common than the ones for a PhD-level function. What

stands out is that the ‘Time to Select’ and ‘Time

Interview to Offer’ are very high for PhD-level

functions.

Table 4: Bachelor, Master, and PhD recruitment scorecard.

Stage 5. Evaluation.

During this stage, the HR professional evaluates the

results of the recruitment process mining analysis to

determine whether the metrics provide meaningful

insights that answer the business questions at hand.

The process model and its statistics are then presented

to and discussed with relevant business experts and

stakeholders. Based on this discussion, it may be

determined that the initial questions have not been

fully answered, or new follow-up questions have

arisen, which may require additional iterations of

2HR

Stage 6. HR Process Improvement.

Once all the business questions have been

satisfactorily answered, or the desired insights have

been obtained, it is possible to take targeted action to

improve process performance. In the context of

recruitment, this could involve focusing on faster

screening times by recruiters, streamlining the

process of scheduling interviews with applicants, and

so on. An essential aspect of process improvement is

conducting additional iterations of analysis. This

involves updating the dataset and repeating the

analysis to obtain the latest results, which can help

evaluate the effectiveness of the actions taken. If the

targeted actions have led to progress, then they have

been successful in improving process performance.

However, if the results are not as expected, additional

actions may be necessary to address the identified

issues. Frequent iterations of analysis on recent data

not only allow for continual improvement of process

performance but also help ensure that the

organization remains in control. This approach helps

to identify potential problems early on and make

appropriate adjustments, ultimately leading to more

effective and efficient business operations.

In general, the results from the case study were

highly valuable for the company's HR department, as

they provided crucial day-to-day insights into the

operational excellence of the recruitment process.

These insights were challenging to obtain and

generate before, highlighting the usefulness of

process mining.

ICSOFT 2023 - 18th International Conference on Software Technologies

650

4 CONCLUSIONS

This paper provides an overview of the PM

2HR

methodology for HR analytics. This process mining

project methodology for HR analytics is a

comprehensive approach to multiple HR decision-

making problems and it is developed based on the

methodology (Van Eck et al., 2015). In the

following, we list the sources of added value of the

process mining technique in the HR context:

 Process mining allows the HR practitioner to

conduct analyses not possible with existing

tools, such as the classical data mining and

statistical tools.

 Process mining analyses the entire collection of

HR event data.

 Process mining allows HR to have a more

effective way of measuring process efficiencies

and inefficiencies as well as simply providing

an overview of the actual HR process.

To evaluate, the PM

2HR

methodology for HR

analytics we applied it on a case study related to

recruitment analysis. Overall, the application of the

methodology was successful as it produces relevant

insights in the recruitment process, with the support

of the HR performance scorecard. A key element in

the methodology is the establishment of the definition

of the HR event log. Another highlight of our research

is the guidance we provide to the HR professional in

assessing to which HR functions process mining

could be used.

To conclude, this paper demonstrates that the use of

process mining in the HR domain creates new

insights that enable better HR-decision making,

particularly in the recruitment and selection process.

We demonstrate that the PM

2HR

methodology for HR

analytics is a comprehensive user guide for process

mining analyses in the HR domain. We also illustrate

that these analyses lead to significant process

improvements. However, we believe that there is still

need for applying and evaluating the PM

2HR

methodology for HR analytics in HR process mining

projects of other organizations and conducting more

case studies.

ACKNOWLEDGEMENTS

We would like to thank W. van der Zanden and M.E.

Savvidi for their contribution to this work.

REFERENCES

Van der Aalst, W.M.P. (2009). Process-Aware Information

Systems: Lessons to Be Learned from Process Mining.

In: Jensen, K., van der Aalst, W.M.P. (eds)

Transactions on Petri Nets and Other Models of

Concurrency II. Lecture Notes in Computer Science,

vol 5460. Springer, Berlin, Heidelberg.

van der Aalst, W.M.P. (2012). Process mining.

Communications of the ACM, 55:76–83.

van der Aalst, W.M.P. (2016). Process mining: Data

science in action. Springer Berlin Heidelberg, 2 edition.

Arias, M. & Saavedra, R. & Marques Samary M., Munoz-

Gama, J. & Sepulveda, Marcos. (2018). Human

resource allocation in business process management

and process mining: A systematic mapping study.

Management Decision. 56. 10.1108/MD-05-2017-

0476.

Chauhan, A., Sharma, S. K., & Tyagi, T. (2011). Role of

HRIS in improving modern HR operations. Review of

Management, 1(2), 58-70.

Christensen, C. M., Hall, T., Dillon, K. & Duncan, D. S.

(2016). Know Your Customers’ “Jobs to Be Done”.

Harvard business review, 94(9), 54–62.

van Eck, M. L., Lu, X., Leemans, S. J., & van der Aalst, W.

M. (2015). PM2: A Process Mining Project

Methodology. In International Conference on

Advanced Information Systems Engineering (pp. 297-

313). Springer, Cham.

Laumer, S., Maier, C., & Eckhardt, A. (2015). The impact

of business process management and applicant tracking

systems on recruiting process performance: an

empirical study. Journal of Business Economics, 85(4),

421-453.

Jooseok, L., Lee, S., Kim, J. & Choi, I. (2018). Dynamic

human resource selection for business process

exceptions. Knowledge and Process Management. 26.

10.1002/kpm.1591.

Marler, J.H. & Boudreau, J.W. (2017) An evidence-based

review of HR Analytics, The International Journal of

Human Resource Management, 28:1, 3-26, DOI:

10.1080/09585192.2016.1244699.

Muenstermann, B., Eckhardt, A., & Weitzel, T. (2010). The

performance impact of business process

standardization: An empirical evaluation of the

recruitment process. Business Process Management

Journal, 16(1), 29–56.

R'bigui, H. & Cho, C.. (2017). The state-of-the-art of

business process mining challenges. International

Journal of Business Process Integration and

Management. 8. 285. 10.1504/IJBPIM.2017.10009731.

A Process Mining Methodology for People Analytics: With a Case Study on Recruitment Analysis

651