Customer Journey Analytics: A Model for Creating Diagnostic

Insights with Process Mining

Daan Weijs and Emiel Caron

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

Keywords: Customer Journey, Touchpoints, Customer Journey Analytics, Data Models, Process Mining.

Abstract: The customer journey is becoming more complex due to digitization of business processes, broadening the

gap between the proposed journey and the journey that is actually experienced by customers. Customer

Journey Analytics (CJA) aims to detect and analyse pain points in the journey in order to improve the customer

experience. This study proposes an extended version of the Customer Journey Mapping (CJM) model, to

measure the impact of different types of touchpoints along the customer journey on customer experience, and

to apply process mining to gain more insight in the gap between proposed and actual journeys. Moreover, this

model is used to develop dedicated CJA based on process mining techniques. A case study on e-commerce

applies the CJM-model in practice and shows how the combination of process mining techniques can answer

the analysis questions that arise in customer journey management.

1 INTRODUCTION

The digitization of businesses increases the

complexity of the customer journey. Customers now

interact with a firm trough many touchpoints and they

expect a seamless experience throughout their

journey with the firm. Companies try to manage the

customer journey to create an optimal experience, but

in practice there are many deviations between the

proposed journey and the journey that is actually

experienced by customers. The large amounts of data

that are being generated at touchpoints open up

opportunities to find out how to provide more value

to customers. This study aims to find out how to

analyse and gain better insights into the impact of

different types of touchpoints along the customer

journey on customer experience, by producing

Customer Journey Analytics (CJA) derived from

process mining.

The customer journey encompasses all inter-

actions a customer has with a firm during the end-to-

end purchase process (Lemon and Verhoef, 2016).

Customer experience is built over an extended period

of time and includes many touchpoints between a

customer and an organization (Zomerdijk and Voss,

2011). A touchpoint can be defined as “an instance of

communication between a customer and a service

provider” (Halvorsrud et al., 2016). All touchpoints

together constitute the customer journey (Zomerdijk

and Voss, 2011) and companies carefully design their

journeys to create an optimal experience. The journey

has a static state which reflects the hypothetical

journey of the service delivery process, i.e., the

expected journey, and a dynamic state that represents

the actual execution of the process (Halvorsrud et al.,

2016), i.e., the actual journey. The shift from offline

to online commerce increases the complexity of

marketing communications, broadening the gap

between the static and the dynamic state. This results

in new types of management questions about the

dynamic state of the journey, that support marketing

decision-making. These questions revolve around

determining the effectiveness and the impact of

different marketing communications on business

outcomes, such as revenue or customer experience.

The data that is generated at online touchpoints

open up opportunities for the visual representation of

the customer journey, i.e., maps, and applying

analytics (Holmlund et al., 2020; Lemon and

Verhoef, 2016), which helps to answer the questions

that arise in customer journey management. Figure 1

presents an overview of the elements of an online

customer journey. A firm creates different types of

advertising touchpoints to attract visitors to its

website. A website typically has a funnel structure

that supports visitors with relevant information at

every stage of the customer journey, moving from

418

Weijs, D. and Caron, E.

Customer Journey Analytics: A Model for Creating Diagnostic Insights with Process Mining.

DOI: 10.5220/0011263900003266

In Proceedings of the 17th International Conference on Software Technologies (ICSOFT 2022), pages 418-424

ISBN: 978-989-758-588-3; ISSN: 2184-2833

Figure 1: Elements of an Online Customer Journey, adapted from Anderl et al., 2016; de Haan et al., 2016; Lemon and

Verhoef, 2016.

Figure 2: Components of the online Customer Journey Mapping (CJM) model, adapted from Bernard and Andritsos (2017).

general interest to more specific consideration of the

actual purchase or download. The post-purchase

phase includes reviews, customer lifetime value and

repurchase rates. Every action that a visitor performs

on the website can be logged. In addition, a website

registers how the user landed on the website. Data

generated at the advertising and website touchpoints

is a rich source for analytical purposes. The goal of

CJA is to enable marketers to detect and analyse pain

points, allowing them to intervene and develop

possible actions to improve customer experience.

The objective of this research is threefold: (1) to

extend the Customer Journey Mapping (CJM) model

of Bernard and Andritsos (2017), to measure the

impact of different types of touchpoints along the

customer journey on customer experience, (2) to

develop dedicated CJA based on specific process

mining techniques, and (3) demonstrate the results of

(1) and (2) on a real case study in e-commerce.

This paper is structured as follows. In Section 2

we propose the extended CJM-model and connect

CJA with process mining techniques. After that, we

apply the model in a case study to produce CJA for

an e-commerce company. Finally, conclusions are

drawn in Section 4.

Customer Journey Analytics: A Model for Creating Diagnostic Insights with Process Mining

419

2 PROCESS MINING FOR CJA

2.1 Extended CJM-model

Process mining aims to discover, monitor and

improve business processes (Aalst, 2012). It can be

used in many different contexts and settings where

processes are executed. One of the main contributions

of process mining is that it uses real event data that is

generated during processes, closing the gap between

assumed process models and the actual execution of

the process. Bernard and Andritsos (2017) proposed

a CJM-model to store CJMs as XML structures and

they illustrated how process mining can be used to

analyse CJMs. While the original model provides a

structured starting point for mapping, the advertising

elements and the website funnel structure (see Figure

1) have to be included as model components to create

a more comprehensive overview of an online

customer journey. In addition, a further mapping

between process mining techniques and dedicated

managerial CJA questions is required to discover the

applicability of the techniques within this domain.

Figure 2 presents the complete CJM-model for

online customer journeys in terms of an UML

(Unified Modeling Language) class diagram. The

CJM-model consists of multiple customer journeys.

Each Journey is performed by a Customer, and a

customer can perform one or many journeys. A

customer journey might include six dimensions,

captured by the aggregations and components of the

Journey class. A Journey is started from a physical

Location (1), performed on a specific Device (2), is

started from a specific channel, i.e., Campaign, as

defined by Anderl et al. (2016) (3), has a Landing

page (4), and a Date (time) dimension (5), that

captures when the journey started. In addition, it

consists of multiple Touchpoints (6) that each have a

timestamp and a descriptive name, a Page dimension

where the touchpoint is encountered, a Stage (i.e.,

pre-purchase, purchase and post-purchase phases), as

defined by Lemon and Verhoef (2016), and an

Experience that can be measured with an emotion,

scale or quote (Bernard and Andritsos, 2017). Here

the model of Bernard and Adritsos (2017) is extended

by adding the Journey – Touchpoint hierarchy

(dimension 6) and dimensions 1 – 4.

The hierarchy of the online CJM-model is

reflected in the import data to make meaningful

analysis with process mining software. The

requirements of an event log - a case id, timestamp

and event - are met in the Touchpoint class. However,

a single flattened event log, i.e., fact table, including

all touchpoints is not enough to capture the structure

of the customer journey. The dimensions that

characterize the journey are important and need to be

included in the data model. This allows for slicing and

dicing the data from multiple perspectives and

enables marketers to detect and analyse pain points,

for example, by comparing journeys from different

campaigns or journeys that started from desktop

versus mobile.

2.2 Mapping Process Mining with CJA

The CJM-model in Figure 2 shows that the customer

journey and process mining can directly be related in

terms of the required data structure. Moreover, we

map classes of process mining techniques to specific

CJA analysis questions. The techniques applied in

process mining should be able to provide meaningful

insights that managers can use to manage the

customer journey. The goal of CJA is to enable

marketers to detect and analyse pain points and

opportunities in the customer journey in order to

develop possible optimization efforts. CJA should

focus on the gap between the proposed journey and

the actual journey as experienced by the customer.

Process mining includes three main classes of

analysis: discovery, conformance checking and

enhancement (Aalst, 2012). Process discovery uses

event log data to create a process model of the actual

execution of the process. Conformance Checking

compares the proposed model with the actual model

to check the process conformance. Enhancement aims

to change or extend the process model, based on the

insights from the other techniques. In addition to the

types of analysis, process mining covers different

perspectives (Aalst, 2016). The control-flow

perspective focuses on the ordering of activities, i.e.,

the control flow. The organizational perspective

focuses on the resources, i.e., actors, people, roles,

departments, that are involved in the process and how

their tasks are related. The case perspective includes

the properties of the cases. It focuses on the individual

characteristics of cases beyond the activities or

resources. Finally, the time perspective focuses on the

timing and frequency of events.

Process mining techniques have shown to be

effective in analysing gaps between proposed and

actual processes by discovering the actual process

model based on event logs (Aalst, 2016). The

objective of process mining is to discover, control and

improve actual processes. These objectives fit the

types of business questions that arise in customer

journey management, where process discovery can be

used to discover the actual customer journeys,

conformance checking to analyse the gap between the

ICSOFT 2022 - 17th International Conference on Software Technologies

420

proposed and the actual journey, and enhancement

could focus on new events and attributes to create

richer analysis. In Table 1, CJA is aligned with

process mining analysis. The table presents an

overview of which mining techniques and

perspectives are used to answer the different types of

analysis questions that arise in customer journey

Table 1: Mapping process mining with CJA.

PM Analysis Perspective CJA questions

Discovery

Control-flow What is the most

common path on the

website?

Case What are the

average events per

case for different

journeys?

Time How quick do

visitors drop out in

the path to

purchase? (Lemon

and Verhoef, 2016)

Conformance

Control-flow Where do visitors

deviate from the

customer journey

map?

Control-flow Where in the

purchase process do

visitors drop out?

(Lemon and

Verhoef, 2016)

Case Was my marketing

campaign

successful?

Time Which types of blog

content are

engaging?

Enhancement

Control-flow Which new events

can be included to

create richer

analysis?

Case Can we cluster

visitor behaviour to

identify customer

segments? (Lemon

and Verhoef, 2016)

Time How does journey

duration deviate for

different product

segments?

https://analytics.google.com

management. The alignment is based on careful

review of the literature on process mining and CJA

(Aalst, 2012; Aalst, 2016; Lemon and Verhoef, 2016)

and validated by case studies (see e.g., Section 3). In

this respect, relevant mining algorithms for CJA are:

Directly-Follows Graphs, Heuristics miner, and

Fuzzy miner, that take into account the frequency of

events, and where the output of the algorithm is

transformed into dedicated CJA.

2.3 Customer Journey Data

The website is the central pillar in online customer

journeys (see Figure 1). Google Analytics

(GA) is a

popular technology for tracking and analytics of

websites. GA uses first-party cookies to track

individual users as they navigate through a website.

The latest version, GA4, tracks user behaviour in an

event structure. Every action that a user performs on

a website – e.g., clicking on a link, reading a blog

page, or playing a video – is called a ‘hit’ and triggers

an event, generating huge amounts of data. All user

behaviour on a website is hierarchically ordered in

three classes: A user has one or more website sessions

that each contain one or more events. The class

structure of GA4 corresponds with the extended

CJM-model as explained in Section 2.1. A user

corresponds to the customer, a session corresponds to

a journey including its dimensions and an event

corresponds to a touchpoint. The data, generated by

GA4 can be mapped directly to the CJM-model of

Figure 2. The end result of this mapping is an actual

customer journey that can be visualized with process

mining techniques.

3 CASE STUDY IN

E-COMMERCE

The case study is conducted for the ABC company

an e-commerce company that sells products via their

own website and via third-party e-tailers. On average

their website receives more than 150 thousand

visitors per month. Their goal is to guide visitors

through the journey to the product page where they

can buy the product or click through to a third-party

e-tailer (i.e., lead generation). The ABC-company

distinguishes touch-points in five categories related to

their website funnel: homepage, blog-post, category-

page, product-page, and checkout. An underlying

data model and (import) data set have been created

The actual name of the company is not disclosed.

https://www.celonis.com

Customer Journey Analytics: A Model for Creating Diagnostic Insights with Process Mining

421

Figure 3: Discovered actual customer journeys for the ABC-company.

based on the CJM-model (see Figure 2) and applied

in the case study. The data is loaded into Celonis

process mining software and subsequently a part of

the CJA questions from Table 1 are used to test and

validate its usability with the software.

Figure 3 shows the discovered process model with

‘fuzzy mining’ and provides insights in the flow of

visitors along the touchpoints in the customer

journey. Most of the journeys start on a blog-post and

the least start on the homepage. More than 50% of the

journeys end directly after viewing a blog-post.

Process discovery with a control-flow perspective

shows the most common path on the website. The

time perspective can be used to find out how quick

visitors drop out in the process. Process discovery

techniques help analyse the gap between the static

and the dynamic state of the customer journey.

Conformance checking and process analytics

techniques measure the performance of different

customer journeys. Table 2 shows the journey

conformance and purchase probability of four

expected journeys, based on the website funnel

depicted in Figure 1. Company ABC proposed four

types of expected journeys, that can start at either one

of the pages. Journey conformance uses the control-

flow perspective and measures the percentage of

journeys that conform to the proposed flow. It

answers the questions where visitors deviate from the

proposed journey and where in the purchase process

visitors drop out, by showing the actual journeys that

visitors have. The purchase probability measures the

probability that a visitor makes a purchase, given that

he followed the proposed journey, using conditional

probability P(A|B). These insights help a company to

measure the impact of different types of touchpoints

along the customer journey towards customer

experience. Table 2 shows the biggest gap between

the expected and the actual journey for the Blog-

Category-Product journey. Only 10.31% conforms to

the proposed journey, meaning that about 90%

follows a different path. This indicates some

bottlenecks in the journey and since the purchase

probability is relatively high, it is valuable to find

areas of improvement. By diving deeper into these

journeys and slicing the data on the 5 perspectives of

a journey, the root causes behind the pain points can

be identified. Two insights were a shifted ratio of

mobile/desktop traffic for these journeys (i.e., the

Device dimension) and a paid search campaign that

led visitors to a wrong blog category (i.e., Campaign

dimension).

Table 2: Conformance Checking and Process Analytics.

Expected

Journey

conformance

Purchase

probability

Homepage 

Category 

Product

46.97% 2.34%

Blog 

Category 

Product

10.31% 2.44%

Category 

Product

46.28% 2.99%

Product - 1.39%

The product page contains five types of

touchpoints that visitors can encounter. In addition to

the product information, these pages include elements

such as a video player or a button where visitors can

ICSOFT 2022 - 17th International Conference on Software Technologies

422

download a file. The interactions on these elements

are measured by custom events. Process analytics

provides insights in what behaviour leads to a

purchase. The conditional probability P(A|B) of each

of these touchpoints has been calculated towards a

purchase or a lead generation. In this case, A is the

purchase and B is the journey that the visitor

followed. The impact on purchase, measures the

probability that the visitor makes a purchase, given

that (s)he followed the proposed journey. Table 3

shows the results. For example, for the review

impression the probability of making a purchase is

2.86% given the fact that the visitor has checked the

reviews of the product. The same calculation is done

for lead generation, where the probability of clicking

through to a third-party e-tailer is 13.59%. Both

measures indicate the effect of a touchpoint. Table 3

shows the values for all journeys, but these values can

be filtered on different product segments, campaigns

or mobile versus desktop journeys to gain more

insight into purchase behaviour under different

circumstances.

Table 3: Process Analytics on the product page.

Touchpoint Purchase

probability

Lead

generation

probability

View image

gallery

2.39% 7.05%

Select product

options

3.81% 11.80%

Check the

reviews

2.86% 13.59%

Play the

product video

3.52% 10.41%

Download a

file

1.49% 5.16%

Based on the findings of the process mining

analysis, some suggestions for improvement can be

made, for example, changing the blog categories for

paid search campaigns and optimizing the website for

mobile devices. The extended CJM-model, with the

Journey – Touchpoint hierarchy and additional

dimensions, allows for these analyses. The insights

provided result from slicing the data on location,

device, campaign and landing page characteristics.

Due to space limitations, only brief examples are

presented here. The case study shows that the

combination of process mining techniques can

answer the questions that arise in customer journey

management. Moreover, another case study has been

conducted in a business-to-business environment,

showing similar results. We are now working on

extending the model’s analytical capabilities to

ensure its generalizability to other web applications.

4 CONCLUSIONS

This paper contributes to the CJA and process mining

research by: (1) extending the CJM-model to measure

the impact of different types of touchpoints on

customer experience, by adding the Journey –

Touchpoint hierarchy with several analytical

dimensions that allow for a richer analysis of the

customer journey, (2) developing dedicated CJA

based on process mining, and (3) demonstrating the

results in an e-commerce case study. It further closes

the gap between CJA and the application of process

mining techniques. The case study applies the

extended CJM-model in practice and shows how the

different process mining techniques can be used to

create insights in the customer journey. The ability to

slice the data from multiple perspectives allows for

new insights that were not possible with the model of

Bernard and Andritsos (2017). In addition, it shows

the application of the theoretical model in practice,

closing the gap between academia and managers.

Further research could focus on applying the model

in a more extensive case study to validate if the

improvement efforts that result from the process

mining analysis indeed help a firm to improve

customer experience. In addition, a more complete

evaluation in comparison with existing CJA

techniques is envisaged.

REFERENCES

Aalst, W. V. D. (2012). Process mining. Communications

of the ACM, 55:76–83.

Aalst, W. V. D. (2016). Process mining: Data science in

action. Springer Berlin Heidelberg, 2 edition.

Anderl, E., Schumann, J. H., and Kunz, W. (2016). Helping

firms reduce complexity in multichannel online data: A

new taxonomy-based approach for customer journeys.

Journal of Retailing, 92:185–203.

Bernard, G. and Andritsos, P. (2017). A process mining

based model for customer journey mapping.

de Haan, E., Wiesel, T., and Pauwels, K. (2016). The

effectiveness of different forms of online advertising

for purchase conversion in a multiple-channel

attribution framework. International Journal of

Research in Marketing, 33:491–507.

Halvorsrud, R., Kvale, K., and Følstad, A. (2016).

Improving service quality through customer journey

analysis. Journal of Service Theory and Practice,

26:840–867.

Customer Journey Analytics: A Model for Creating Diagnostic Insights with Process Mining

423

Holmlund, M., Vaerenbergh, Y. V., Ciuchita, R., Ravald,

A., Sarantopoulos, P., Ordenes, F. V., and Zaki, M.

(2020). Customer experience management in the age of

big data analytics: A strategic framework. Journal of

Business Research, 116:356–365.

Lemon, K. N. and Verhoef, P. C. (2016). Understanding

customer experience throughout the customer journey.

Journal of Marketing, 80:69–96.

Zomerdijk, L. G. and Voss, C. A. (2011). Nsd processes and

practices in experiential services. Product Development

& Management Association, 28:63–80.

ICSOFT 2022 - 17th International Conference on Software Technologies

424