=Paper=
{{Paper
|id=Vol-1920/BPM_2017_paper_172
|storemode=property
|title=CJM-ex: Goal-oriented Exploration of Customer Journey Maps using Event Logs and Data Analytics
|pdfUrl=https://ceur-ws.org/Vol-1920/BPM_2017_paper_172.pdf
|volume=Vol-1920
|authors=Gaël Bernard,Periklis Andritsos
|dblpUrl=https://dblp.org/rec/conf/bpm/BernardA17
}}
==CJM-ex: Goal-oriented Exploration of Customer Journey Maps using Event Logs and Data Analytics==
https://ceur-ws.org/Vol-1920/BPM_2017_paper_172.pdf
CJM-ex: Goal-oriented Exploration of Customer
Journey Maps using Event Logs and Data
Analytics
Gaël Bernard1 and Periklis Andritsos2
1
University of Lausanne, Faculty of Business and Economics (HEC), Switzerland
gael.bernard@unil.ch
2
University of Toronto, Faculty of Information, Canada
periklis.andritsos@utoronto.ca
Abstract. Customer Journey Mapping (CJM), is an emerging area of
research tackling issues related to customer behavior and user trajectories
when consuming a service. The increasing complexity of the service in-
dustry makes this type of tools popular amongst practitioners. However,
to date, it is not clear how a CJM can be used to depict hundreds or
thousands of customer journeys. Inspired by process discovery techniques
– borrowed from Process Mining – we present CJM-explorer (CJM-ex).
CJM-ex is a web interface that uses hierarchical clustering and statistical
indexes to allow interactive navigation, with or without a-priori informa-
tion, through numerous journeys stored in standard event log formats.
The exploration of the underlying journeys can be done in the whole set
of data available or driven by user goals in order to examine events and
patterns in specific areas of interest.
Keywords: customer journey mapping, process mining, customer jour-
ney analytics, hierarchical clustering, sequence mining
1 Motivations
In order to deliver great service, companies need to have an understanding
of the quality of customer experience at an end-to-end level, [5,7]. The ever
growing amount of services offered to users for consumption has made the ability
to understand their behavior very important, [8]. Similarly important is the
knowledge extracted by the increasing number of ways organizations interact
with their customers; e.g., a customer might visit a physical store, purchase
a product online, and provide feedback on social media. As a response, new
customer-centric approaches have surfaced. The customer journey map (CJM)
is an example of such new techniques. CJMs allow for better understanding
of a customer’s end-to-end experience when using a service by mapping any
interactions with the company (called touchpoints) on a map that ultimately
contributes to better understanding and serving customer needs.
In our previous work, [4], we discussed the benefit of bringing customer
journey mapping and process mining together by proposing a CJM model that
can be used with process mining techniques. More specifically, CJM-ex aims at
providing a solution to explore numerous customer journeys at the same time.
Similar to discovery techniques used in process mining, our algorithm takes event
logs as input, without using any a-priori information, [1]. However, instead of
outputting a business process model (BPM), we display the journeys onto a
CJM. For a complete definition of process mining and event logs, the reader is
referred to the well known Process Mining Manifesto, [2]. Visualizing event logs
on CJMs, instead of BPMs, exhibits two interesting features. First, CJMs focus
more on personal customer activities (e.g., by incorporating customer emotions),
rather than the “internally-focus problem-solving approach” of BPMs, [3]. Second,
contrary to BPMs, CJMs can incorporate customer journeys that are deemed
exceptional behaviors, rather than removing them to increase model readability.
Despite these interesting features, representing many customer journeys onto
a CJM in an intelligible manner remains a challenge. Current research typically
limits the number of journeys to be compared to less than ten, making the
overall process relatively straightforward. However, we argue that companies in
the service industries tend to deal with hundreds or thousands of journeys. To
overcome this challenge and identify different areas of interest, a hierarchical
clustering algorithm is employed to segment the original data. The hierarchical
nature allows for a top-down navigation of automatically generated groups of
similar journeys. Once the clusters are formed, CJM-ex is able to leverage the
contextual information that comes along a typical customer journey such as
the customers’ characteristics, or the emotions, [4]. It does so in two different
ways. First, we employ statistical indexes in order to explain why the different
clusters were generated. Second, we let the users define their own exploration
goals, making CJM-ex the first goal-oriented tool that allows analysts to set
a-priori goals to guide their journey exploration.
We implemented CJM-ex, which aims to: 1) show how numerous event logs
can be displayed onto CJMs; and 2) let users navigate into these journeys. The
next section introduces our tool, while the third section highlights the key parts
of its technical implementation. We conclude by providing an outlook.
2 CJM-ex
Representative
Journey …
abz …
The main objective of CJM-ex is to let users upload
acbz …
and explore their own dataset using a customer journey aecbz acgbz
map layout. To limit the number of journeys displayed Actual Journey
on the same CJM and allow their intuitive exploration,
we took a hierarchical clustering approach, as illus- Fig. 1. Hierarchical clus-
tering of journeys
trated in Fig. 1. Each letter represents an activity and
the tree is built bottom up by merging the activities that are most similar at
each iteration. By default, our hierarchical algorithm uses a proximity measure
that takes into account the order of activities and is a variant of the Jaccard
similarity based on shingles, [6]. In our clustering tree (or dendrogram), the
journeys seen in the event logs are at the leaf level and as they get merged they
form “representative” journeys. A representative journey is a single pattern of
activities whose purpose is to summarize the patterns contained in a cluster.
Because the representative journeys at the top of the tree summarize many –
potentially distant – journeys, it will tend to show only few activities shared
by many. In contrast, the representative journeys closer to the leafs will show
more details. Borrowing a cartographic metaphor from [1], the first layers show
general patterns and hide less important activities – like a world map would
omit small cities. However, our application allows a drill-down into ‘countries’ of
interest (i.e., pattern of activities), which would redirect to new CJMs where the
previously hidden ‘cities’ (i.e., omitted activities) will be shown.
CJM-ex is accessible at http://
customer-journey.unil.ch, where we 1 2
also provide a screencast explain-
ing its usage. The screenshot visi-
ble in Fig. 2 points out three views
that are available to navigate the
clusters. Each of these views full-
fills specific objectives. First, the 3
CJM view Ê shows journeys that
are in the same cluster. This repre-
sentation allows to easily compare
the pattern of activities. Second,
the tree Ë displays the hierarchi- Fig. 2. Interface pointing to three views: Ê CJM,
cal structure of the journey clus- Ë tree and, Ì textual boxes.
ters – useful in providing a holistic
view of the clusters and where we currently are. Third, a box per cluster Ì
provides a convenient means to display statistical indexes that we named “salient
characteristics”. The salient characteristics is the top 5 results of a chi-square test
applied on all the contextual information. For instance, if at a global level (the
entire dataset) the number of women is equal to the number of men, it might
be surprising to find a cluster with large majority of women. Therefore, this
information might come up as one of the top 5 salient characteristics.
Moreover, the user might be interested in specific
characteristics occurring during the journey. For this
reason, we allow user-defined goals. For instance, one
might be interested in journeys that started by the
activity “attending class” experienced by young people.
The top part À of Fig. 3 displays the settings, while the
bottom part Á shows that some branches of the tree
are interesting with regards to the goal (red “hot” area
at the top). Hence, our application allows navigation
without using any a-priori information, but also setting Fig. 3. À goal settings, Á
navigation goals, and guidance by the resulting colors. impact on the tree.
Finally, when moving from one view to the others, the three views are updated
synchronously, allowing a smooth exploration amongst journeys.
1,2,3,4,5,6,7,8,9 ab
1. Hierarchical 2. Recursive 1,2,3,4 5,6 7,8,9
3. Finding abc abxy abz
clustering 1 2 3 4 5 6 7 8 9
cuts 1,2 3,4
Representative abcd abce
Fig. 5. Data preprocessing phase
3 Implementation
CJM-ex is build around four main elements: 1) a web interface; 2) the XES-parser;
3) Hcluster; and 4) a data warehouse. We will describe the main parameters and
choices we made for each of them.
Web interface. The web interface leverages bootstrap3 , jquery4 and d3js5
to provide a user-friendly interface to upload and navigate journeys. Both the
CJM view and the tree view are implemented in d3js. The CJM view is our own
implementation, while the tree is an adaptation of existing code6 .
XES Parser. CJM-ex works with event logs. More specifically, we leverage
the XES (eXtensible Event Stream) standard born within the process mining
taskforce. The XES Parser is a Java implementation that encapsulates the
OpenXES library7 to parse XES file. Use of this open source software ensures
that our application is strictly compatible with the XES standard.
Hcluster. Hcluster is a python implemen-
tation containing the three steps illustrated in
Fig. 5. The first one is the hierarchical cluster-
ing implemented using Scipy8 . Two parameters
should be provided as inputs: the distance mea-
sure between event sequences and the methods
for calculating the distance between clusters.
They can both be chosen by the user when up-
Fig. 4. Preview of the parameters
loading a dataset (see Fig. 4). While further when uploading a dataset
research is needed to understand why these dif-
ferences exist, it seems that using shingles, [6], as a distance metric provides a
more intuitive way to navigate through journeys. Once the hierarchical cluster is
formed, the next step consists of cutting the clustering to form layers. A layer is
a set of predetermined number of journeys that will be grouped together and will
ultimately appear on the same CJM. To achieve this, we developed an algorithm
that recursively cuts the dendrogram returned by Scipy. A small number of
journeys will lead to a simple CJM that is easy to visualize, but in more complex
tree structures (i.e., trees with larger height). Finally, the last step consists of
finding the representative journey using a frequent sequences mining algorithm9 .
Data Warehouse. Each dataset is saved in its own database schema designed
as a star schema. Due to lack of space, we do not include its full schema, which
3
http://getbootstrap.com/
4
https://jquery.com/
5
https://d3js.org/
6
http://bl.ocks.org/robschmuecker/7880033
7
http://www.XES-standard.org/openXES/start
8
https://docs.scipy.org/doc/scipy/reference/generated/scipy.cluster.hierarchy.linkage.html
9
https://github.com/bartdag/pymining/blob/master/pymining/seqmining.py
stores all the information required to use the application (e.g., clusters, journeys,
events) as well as some precomputations. For instance, we count the number of
occurences for each characteristic at each cluster, so the goals and the salient
characteristics can be retrieved quickly.
Altogether, the parameters visible in Fig. 4 allow users to explore the exact
same dataset from different perspectives.
4 Discussion and outlook
With CJM-ex, we have demonstrated that many journeys can be displayed onto
CJM in an intelligible and efficient manner, offering an alternative to a BPM
representation. As Gartner highlighted, CJMs should be used to complement,
but not to replace, BPMs, [9]. By leveraging a standard born within the process
mining taskforce (i.e., XES) and by mimicking a typical process mining activity
(i.e., discovery), we bring these techniques closer together. However, further
research is required to fully understand how they can complement each other.
This is a call to the process mining and business process management communities
to consider CJMs as an integral part of the process management toolkit.
References
1. van der Aalst, W.: Process Mining: Data Science in Action. Springer (2016)
2. van der Aalst, W., Adriansyah, A., De Medeiros, A.K.A., Arcieri, F., Baier, T.,
Blickle, T., Bose, J.C., van den Brand, P., Brandtjen, R., Buijs, J., et al.: Process
mining manifesto. In: International Conference on Business Process Management.
pp. 169–194. Springer (2011)
3. van den Bergh, J., Işik, O., Viaene, S., Helsen, E.: Re-positioning business process
management: Exploring key capabilities for successful business transformation (2016)
4. Bernard, G., Andritsos, P.: A process mining based model for customer journey
mapping. In: Proceedings of the Forum and Doctoral Consortium Papers Presented
at the 29th International Conference on Advanced Information Systems Engineering
(CAiSE 2017) (2017)
5. Bernard, G., Boillat, T., Legner, C., Andritsos, P.: When sales meet process mining:
A scientific approach to sales process and performance management. In: International
Conference on Information Systems (2016)
6. Broder, A.Z., Glassman, S.C., Manasse, M.S., Zweig, G.: Syntactic clustering of the
web. Computer Networks and ISDN Systems 29(8-13), 1157–1166 (1997)
7. Haugstveit, I.M., Halvorsrud, R., Karahasanovic, A.: Supporting redesign of c2c
services through customer journey mapping. In: Service Design Geographies. Pro-
ceedings of the ServDes. 2016 Conference. pp. 215–227. No. 125, Linköping University
Electronic Press (2016)
8. Lemon, K.N., Verhoef, P.C.: Understanding customer experience throughout the
customer journey. Journal of Marketing pp. jm–15 (2016)
9. Olding, E., Cantara, M., Robertson, B., Dunie, R., Huang, O., Searle, S.: Pre-
dicts 2016: Business transformation and process management bridge the strategy-
toexecution gap. Tech. rep., Gartner (November 2015), https://www.gartner.com/
doc/3173020/predicts--business-transformation-process