Case Study

Developing Behavioral Personas Using Cluster Analysis

Personalized user journeys by segmenting users based on their activity patterns.

The Problem

For many companies, there is an increasingly short time between developing a new feature and releasing it to users. A product that is continuously improved has become the new norm for many SaaS companies. However, it doesn't make much difference to get features into the hands of users if they are unable to learn how to use them to get value.

It was the case with this Digital Signage Platform and their ever-increasing number of available features, exposing newly registered users to many decisions right from the start, without providing a clear path to success.

This situation soon became a problem, with users trying out many things and abandoning the product shortly after.

The current approach of presenting users with just industry-based content templates and disregarding the features until later was not practical, as they would have to figure out by themselves what to do with the content and be faced with many other decisions, like playlist configurations, dynamic content options and more. We needed an approach where users could accomplish something valuable for them during their first session and be motivated to return and adopt the product.

The Research

For this research project, I wanted to focus on current users' in-app activity and build segments around that. Too often when we look at new users it's hard to identify what is actually working in the long term, what ended up retaining a user, so figuring out what current users do will help to uncover the core experience that users already find value in and replicate their path to success.

I wanted to answer the following questions: How are users segmented based on activity and spending? Which features could add the most value to each segment?

Once I had this I could create a set of Behavioral Personas and use them as the base for the user journey improvements.

The Method

To understand how these groups differ in terms of their activity, I exported some raw event data from Google Analytics and BigQuery. In order to analyze I've prepared it and grouped all events related to content creation and events related to management and calculated their percentages based on total events.

I've used a k-means clustering algorithm to determine which users belong to each cluster based on how much they spend and the ratio of content vs. management events during a 30 day period.


Attribute Cluster 0 Cluster 1 Cluster 2 Cluster 3
Avg. Spending $2424.44 $15.60 $22.96 $491.7
Avg. Users 2.7 2.3 2.3 2.0
% Content 30 Day activity 45.5% 92.4% 58.6% 28.1%
% Management 30 Day activity 37.6% 4.4% 24.54% 62%
% Sessions w/ both activities 33% 7% 47% 46%
Behavioral Personas

The next step was to take the analysis and turn it into behavioral personas. Contrary to traditional personas defined by demographics or interests, behavioral personas segments users out by patterns in how they use a product.

From the cluster analysis, we found some groups that were representative of different behavior patterns.

At first glance, it seems that the Cluster 0 group is the ideal customer, but it is much harder to acquire them and improbable that we can make most users behave similarly.

It is clear that Cluster 1 might not be the perfect customer for our business, they spend more time creating content but don't get much value from managing it, so they spend less overall.

The opportunity then is in Cluster 2 and Cluster 3. It seems that Content and Management usage are balanced, the perceived value increases and consequently leads to higher spending. If both Cluster 2 and Cluster 3 had clear ways to balance their activity and become better, would they also find more value?

Persona Profiles

Bridging the gap between creators and curators

This user knows how to create content and will spend a majority of time uploading this new content to a new playlist. They have many displays but show the same playlist in all of them. Most sessions are task-oriented, with no clear goal completed by the end of it. While this user fits the criteria for a heavy user, there are no indications of the user becoming an expert on digital signage and will likely churn in the future.

This user knows how to take advantage of digital signage, and has competence in creating a digital narrative by using a variety of content types and considering the sequence and optimal placing for it on a playlist. Sessions are goal oriented. They know creating all of their content is time-consuming, so they leverage online sources, this makes for fresh and relevant content distributed across their displays.


Personalized User Journey

With a clear understanding of who our priority personas are, we are ready to revamp the user journey to deliver highly personalized and relevant experiences.

Introduce goals, not features

Goals are a great way to help users explore and understand the value of features in a context related to their needs.

Learning through goal-oriented flows

Introducing users to features while they get something done helps them learn and become digital signage experts.

Relevant Content Suggestions

All that is needed to send out communications, organize internal events and share thoughts and ideas on the workplace.

Rule-Based Playlist Configuration

A rule-based playlist optimized for minimum maintaince effort and instant accommodation of new content.

Optimize for what users are already doing

Saving time on routine tasks can leave users with extra room to explore other features. By managing their brand assets, users can add new content faster.

Custom templates with zero configuration

Not only can users create new content easily, all available templates will be customized from the get-go.

Time-Saving Features

Display groups share the same configuration and are much easier to monitor and make changes to. A new display can be added and activated simply by assigning it to a group.