Creating Archetypes using quantitative and qualitative data

Case Study

7 min readJan 17, 2021

Overview / Methods

Timeline: 6 months
Team: Service Designer 🙋🏼‍♀️, CX Consultant, Data Scientist, Business Analyst
Materials & Tools: Google sheets, G-Slides, Miro.
Research Techniques: User interviews, Customer Journey, Affinity diagram, Data Clustering, Desk research (NPS, Surveys)

The Challenge

It’s possible to identify groups of users with data triangulation? (data-based analysis and qualitative variables) 🕵🏼‍♀️

A bit of context

A payment platform identifies the necessity to have a better understanding of their client’s segments. (It might seem an obvious necessity, but that’s why I need to talk about the context of this company)

This is a payments company with many years of trajectory, and the strategy was shifting from a marketing segmentation to a more customer-centric product development.

Currently, this company has a captive and diverse B2B market. Traditional segmentation is no longer helping the company to develop new and innovative products. So the expectations of this work was to provide teams (business, marketing, product) a new way to see the needs and motivations of their clients in order to create products that help shops and businesses to grow.

The process

Empathize with the users

The method selected for understanding the user's needs and motivations was “In deep interviews” in context.

Choosing the right sample was also challenging because the universe of clients is gigantic. So, what we did was to choose a number of companies in each commercial segment (Small businesses, Medium, Big, and corporations) and sales channel (in-store or e-commerce) In order to identify behaviors and needs according to the company’s turnover and their predominant sales channel.

Planning the research

In order to execute properly that quantity of interviews, we require a mini phase of preparation. Create scripts, canvases, manage the recruitment (this was not an easy thing), and create in advance the format of documenting the information. This allows us to execute interviews simultaneously aligned with the research goals.

As the user was in context (serving their clients), we must develop concise scripts and easy note-taking systems. We use a board with the different phases of the service and try to use it to trigger the conversation.

At the end of the empathize phase, we develop 23 in deep interviews

The hidden step: Systems for documenting

In order to analyze qualitative data in an agile way, an important part of the process is to create a system that allows us to document data from interviews in an organized way.

We could have used a friendlier software for this, like miró or mural, but after a quick overview, google sheets was the best way for a 4 person team to document and categorize properly. This process was key to find patterns and relevant data along the way.

Empathize with the customer service force

A crucial part of the service happens line in line with the customer support staff. Employees behind the phones in call centers, executives behind 1:1 meetings with their clients, and other support services that emerge in the product ecosystem. (Like third party providers that were born as a necessity to solve some problems that this payment processor was not solving).

We complement the understanding phase with interviews, shadowing, and workshops with the staff. In order to gather relevant information about the main pain-points the client was experiencing according to their point of view.

We asked employees, Who are their customers? How are they? how they behave?

Co-create and nurture the archetypes analysis with customer service staffing

A crucial part of the service happens line in line with the customer support staff.

Now is time to analyze the information

This is definitely the most fun and challenging step. Because in record time we must find patterns of information in the different variables researched.

Those variables allow us to understand limits and behaviors, and make our first hypothesis of where an archetype begins and which characteristics are the most relevant to include.

After the analysis, we proceed to build the first iteration of the archetypes founded in qualitative research.

Step 4: Identify key variables of the qualitative archetypes and match them with behavioral data in the database of the company.

What we aim to answer was:

Can we identify behavioral patterns clustering different data from databases of the company?

⚠️ This was a crucial step, but we learned that the success of using databases depends on how clean, organized, and updated this data is.

The process begins by identifying behavioral variables that we could complement with the client’s data storage in the company’s systems.

For example, a qualitative variable was “how the client manages problem-solving”. In the interviews, we identify clients that require complete assistance in order to solve a technical problem and others that are pretty comfortable with self-manage and dream to resolve everything with zero interaction with the company. After that analysis, we wonder, will there exist some data that allow us to better understand that behavior?

Variables that might give us more information:

Quantity of calls to contact center
Quantity of claims processed
Preferred channel to seek assistance and quantity of visits per channel

And after a mathematical and statistical process developed for the data scientist in our team, we could nurture the archetype with more information, and make an estimate calculation of how many are there.

Something interesting is that we had a data ID that helped us math the clients interviewed with the data storage in the databases, (turnover, predominant sales channel, preferred customer service channel, the number of services hired, the antiquity of the business, etc.) that allow us to make data triangulation to enhance the understanding.

Yep, it seems like a mission impossible, but for this company was very important to understand how many clients were related to some behaviors in order to dimension business opportunities, prioritize innovation initiatives, and explore new customer segments with growth opportunities

⚠️ There is not a linear process to find results

After running the statistical model with different variables, the team identifies clusters of companies associated with behaviors from each archetype. Which allows to calculate dimensions of each archetype and quantify representativeness in the client’s universe.

This was not a perfect process, the representativeness does not have a high level of precision, in fact, some of the variables that we aim to match wasn’t on the databases or were outdated, that is why this representativeness was “estimated” and we were ok with that.

The moment of transforming the data in a digestible and actionable format

In this phase is where my instructional designer comes out. In order to do this “archetype” toolkit legible and usable, we must take principles of editorial design, wireframing information architecture, and yep, all that in a google slide. (because really we must use the tools that the client can work and feel comfortable)

We conceptualized some of the main characteristics of the archetypes and with help from a very talented illustrator, we created some visuals that triggered creativity among teams and helped empathize with each archetype.

The Impact

How the company used the archetypes

What was the real impact in the company of this new way to understand their clients?

Once we finished this project, it was very satisfying being able to see in several offices and among teams this “archetype toolkit” in action.

This client’s understanding triggered the creation of new innovation initiatives, ideation processes, and prioritization of new projects.
It was possible for the company to identify the growth potential of some groups of clients, which representative was considered irrelevant before this study.
Designers, managers, product teams now have a new way to understand their clients, which allows them to empathize with their necessities and unleash creative thinking.

The good, the bad, and the ugly

There is not a happy ending with shitty 💩 moments

If I had to do this again, I think it is very important to make a diagnosis of the databases of the company. In order to have the right expectation of the accuracy of the results.
Sample selection must not be taken for granted. All the stakeholders involved must be aligned from the beginning with the sampling strategy. This avoids future questions like “we need to interview more users, the insights are not conclusive”
Do not underestimate the systems and documenting processes. This can be the difference between 3 days of analysis and 2 weeks rambling through the data.
Take pictures of the research activities while they occur, before it is too late and you don’t have any record.
Not having any key stakeholder of the company with the research team in some of the interviews was a bad decision. If I could do this again, I wouldn't let outside this process the people who are going to create and make product decisions. This generates a higher level of empathy and certainty of the insights.

This case represents the huge potential of data triangulation, and how qualitative data and quantitative data can strengthen the knowledge around any subject.

Peace out✌🏼⚡️