Want to understand your user base? Try this user modeling approach

By Valentina Strachan (opens in new tab)

Personas, Customer Segments, and Role-Based Models are all approaches that user researchers employ to represent our insights about types of users and their characteristics. The question of which representation is best is one that’s consistently asked and revisited within our discipline.  

First, why build a user model (UM) at all? To put it simply, discussions at the customer level bring product teams together; user models synthesize and summarize customer information to facilitate those discussions. We have moved from personas to UMs because personas reduce customers to an average that doesn’t exist (opens in new tab). UMs allow for a broader representation of a range of customers.  

To help drive consistency in this area, I led an effort a few years ago to unify a different division’s user models. We started by asking a foundational question: “What goals are UX professionals and product team members trying to achieve with the help of customer type representations?” The answer was one with which all the stakeholders agreed.  

The answer: Main insights  

The three goals for a UM that were rated highest across all respondents were related to why we are representing users in the first place. Our model needed to: 

• Generate empathy: Teams wanted to use the UM to “put them in the users’ shoes” (e.g., having an understanding of users’ demographics/characteristics, pain points, goals, and scenarios when conceptualizing new solutions or to role-play when walking through designs). 
• Be measurable: Which included two components, (1) the user types themselves exist and are accurate – validated through customer feedback and statistical data, and (2) we track the success of our product for each relevant user type. 
• Be agile and adaptive: UX professionals and product team members wanted to quickly perceive, capture, and integrate insights to evolve the UM, so the model stays up-to-date and mirrors the real world. 

The next four highest priority goals had to do with what information the UM portrays. Most current user models don’t do a great job with the majority of this type of info. Team members prioritized: 

• An easy way to contact users: This was one of the most insightful / surprise findings. User models hardly ever give you a way to speak to a user, yet that was the #1 request for specific content to include in a particular role: a link to reach out to this type of user with a question.
• Short labels as a common language/shorthand (e.g., role name) for all the data/information related to one user type, to improve our communication around our users.
• Behavior-based models, or modeling customers based on their behaviors with our products (e.g., through telemetry or screen captures): people with similar behaviors cluster together to define a unique role.
• A comprehensive landscape: The UM provides an understanding of the customer type’s journey across our products and beyond, for a holistic view. 

There were also some goals that were highly rated by individual disciplines. For example: 

• Inform design: The UM provides info about lower-level, feature-specific interactions to help make design decisions. (Ranked especially high with Designers.) 
• Onboarding/refresher: The UM provides a fast way to onboard onto a new product or feature area by providing a digestible synopsis of what we know about the customer type for each product/feature area. (Ranked especially high with PMs and Engineering.) 
• Map user types to division products: An overview of our customers, cross-product scenarios, all-up workflows, and which products are related to which customer type. (Ranked especially high with Leads/Managers.) 
• Screener criteria: Having a screener associated with each customer type (for consistency in recruiting participants to match that segment). (Ranked especially high with Researchers.) 

Impact and recommendations 

Based on the learnings above, we came up with a set of UM design recommendations. Some important approaches included: 

• Single name and photo are an important part of using UM as a communication tool – even if we share multiple sample case studies related to each. 
• Enable direct and indirect contact with the user types – indirect by showing insights related to the segment & direct by surfacing ways of contacting that type of user. 
• Every user type and every insight listed about them is backed by UX research and large-scale data. None of the information is made up or fake (as was the case for personas). Ranges and variability in responses are shown.
• Failed hypotheses are celebrated and visible: E.g., if a role that was hypothesized turned out not to exist in the real world, that knowledge is still valuable.
• Product team members can be active participants in further curating the user model – if they’re allowed to ask questions about the user roles (and perhaps even share their own evidence).
• Include usage telemetry insights about individual user types’ behaviors within and across our products.
• Provide the user journey for each user type, not only within a product, but across products as well. 

While these insights and results were specific to my division at a particular point in time, the methodology we employed in developing our user model can be repeated as our users change and as our needs and goals evolve. By sharing the thought process, my hope is that it will enlighten your efforts to understand your user base.  

What do you think? How has your team used user models? Will these ideas help enhance your work? What would you change or add to these recommendations? Tweet us your thoughts @MicrosoftRI (opens in new tab) or follow us on Facebook (opens in new tab) and join the conversation.

Valentina (Grigoreanu) Strachan is a Senior Design Research Lead. She has a PhD in Computer Science with a focus on Human-Computer Interaction (HCI). In her 10 years as an HCI and UX Researcher, she has conducted in-depth research on a variety of software products for end-user programmers, IT professionals, information workers, software developers, and now consumers. Valentina looks for ways to push the UX Research discipline forward. One common thread across her projects has been to use Artificial Intelligence to inform HCI/UX Research in all phases (e.g., sequential pattern mining and cluster analysis to analyze clickstream data, a conversational agent to improve product team members’ customer conversations). She has also created new research methodologies to maximize the impact of research findings on products (e.g., the Informal Cognitive Method, the Research on Demand toolkit). Valentina has more than 20 publications based on this research, which she has presented at leading HCI conferences such as ACM’s CHI, and IEEE’s VL/HCC, and IFIP’s INTERACT.