Understanding the people who will ultimately engage with a product or service provides the foundation for user experience design. Modeling those people and segmenting our models into meaningful groups lets us explore different clusters of needs, then address our solutions to meeting the needs of people belonging to specific clusters.
Audience segmentation models come in many shapes and sizes. So far, the practice of UX design has focused primarily on the persona as the model of choice. This article explores alternative ways of segmenting audiences and the design research we need to derive each type of model.
Champion Advertisement
Continue Reading…
Personas
In April 2009, I wrote a column for UXmatters titled “User Research for Personas and Other Audience Models,” in which I focused primarily on the research techniques we can use to generate personas, a particular type of audience segmentation model. Personas are archetypal representations of audience segments, or user types, which describe user characteristics that lead to different needs and behaviors. Where the characteristics of users overlap, we build up an archetype, or persona, that represents those users collectively. Where the characteristics of users differ, we must create other personas that represent different audience segments.
However, the discussion around personas tends to focus more on process and communication rather than the type of audience segmentation we’re undertaking. So let’s take a look at some different types of audience segmentation models and how we might go about deriving each of them—that is, what we’re looking for and what research to do.
Audience Segmentation Models
Designing a Web site or application, product, or service for a large, heterogeneous audience can be a daunting task. Typically, we know some members of such an audience are more important to our company than others. To help us design for our most important audience segments, we need some methods of identifying and representing them.
There are many different methods we can use to think about our audience segments, focusing on differences that can help us prioritize and design the features that best meet the needs of each. The type of segmentation we should choose depends on what characteristics would elicit the most meaningful segments.
I’ve already touched on personas and discussed the type of segmentation they produce, but there are many other useful segmentation models, like the following:
market segmentation—Within the business and marketing communities, this has been the most commonly used type of audience segmentation for decades.
experience lifecycle—This segmentation model shows the end-to-end lifecycle of the customer experience.
mental models—This model comprises an affinity diagram of user behaviors surrounding a particular topic.
capability level—This segmentation model indicates the stages of capability our audiences go through over time.
mood—This type of segmentation draws on a concept Will Evans put forward, using mood maps.
game-play style—While this method of segmentation is somewhat specific to the world of online games, this type of segmentation takes into account the way players want to play games—solo or multiplayer.
Market Segmentation
A market segment is a group of people or organizations that share one or more characteristics, causing them to have similar product or service needs. A true market segment meets all of the following criteria:
It is distinct from other segments. Different segments have different needs.
It is homogeneous within the segment. People belonging to a market segment exhibit common needs.
It responds similarly to a market stimulus.
It can be reached by a market intervention.
The term market segmentation is also used when a company divides consumers with identical product or service needs into groups, so they can charge them different amounts of money.
This description of a market segment applies equally well to the other segmentation types I’ll discuss later—particularly the needs for a segment to be distinct from other segments and for all members of a segment to exhibit common characteristics, needs, and behaviors.
Surveys and other forms of analytic data generally drive market segments. Such data might include information about purchasing intent and behavior, demographic information, and frequency of activity in a problem domain.
The primary research tools for market segmentation are, therefore, surveys, interviews—usually closed-question surveys—and focus groups. The analysis process focuses on identifying clusters of characteristics that define distinct segments of an audience—tending heavily toward the multivariate statistical forms of analysis. However, it is possible to derive market segments using fairly simple tools like Excel—though this capability is advanced for such software.
Experience Lifecycle
Experience lifecycle is a generic term that represents the start-to-finish series of interactions a customer has with an organization. For example, LEGO uses an experience wheel like that shown on Customer Experience Matters, depicting the end-to-end experience of a frequent flyer traveling to New York from London.
With this form of segmentation, the goal is to target the design of a product or service to meet the needs of consumers at different stages in their experience. On a recent project that I described in an article on the Johnny Holland Magazine, I broke down the accommodation lifecycle for a backpacker during his or her travels. Different lifecycle stages presented very different sets of requirements and needs for the same person.
You can build experience lifecycles based on a wide range of design research techniques, including observational research, interviews, surveys, analytics, purchasing behavior, and ethnographic studies. The aim here is to understand both the overall structure of the experience—as a lifecycle comprising stages with certain experiential characteristics—and the component tasks of each stage in the lifecycle.
Mental Models
Mental models provide an excellent way of understanding how users approach a context for which we’re designing a product. They let us gain an understanding of collections of activities that make sense to a person—and communicate that understanding to others.
A range of observational research techniques can drive mental models, as well as interviews and, occasionally, focus groups. The key determinant is that the research technique must provide qualitative insights into the ways people approach a problem space, as well as how they cluster tasks into activities and activities into an overall experience.
Capability Level
People can have very different requirements during their early engagement with a product or service versus the requirements they may have once they’ve become familiar with it, then, subsequently, an expert. This is a common requirement for many systems. A user’s early engagement with a system such as a game or social network often requires a great deal more guidance and hand-holding than is necessary once the user has overcome his initial learning hurdles.
To understand the needs of novice versus expert users, it may be necessary to undertake research that is more akin to alpha testing, drawing on techniques from education and technical communication for the solution design. For all its faults, Basecamp handles this progression well, incorporating instructional material into the user interface to help novice users over their initial exposure to the service that disappears over time with use.
Understanding the needs of users with different levels of expertise can involve an analysis of support inquiries. After reviewing support tickets for recurring themes, you can address the problems users are encountering through iterative design and implementation of your solutions.
Mood
In a recent article, “Design Ethnography & Mood Maps,” Will Evans proposed the use of mood maps as a way of describing the various states people go through as they undertake an activity. We can take this approach and modify it somewhat to create a segmentation model that is based on a person’s mood—or frame of mind. For example, an insurance company with an informal brand language may want to use a different style of error message in its online quoting system from the one it uses for, say, its claims application for life insurance policyholders. A segmentation that is based on mood can help highlight such distinctions and remind the design team to take them into consideration. Here is an excerpt from the article that describes mood mapping:
“The [mood map] describes the emotional ups and downs identified by study participants as part of the design exercise conducted during in-home visits with participants. Note that the location of the study is less relevant than the importance of observing the participants in the most likely context in which they will engage in their experience with the brand’s product or service. During the exercise, participants are asked to name each of the phases they went through, from framing their problem through exploration and finally (hopefully) problem solving, and to then assign a corresponding emotion to each phase.”—Will Evans
Contextual studies are key to mood mapping. When seeking information about a person’s frame of mind or mood, neither surveys, interviews, or focus groups, nor Web analytics or other quantitative methods can provide the insights that are available to an observer watching a person as he goes about a task.
Game-Play Style
Joe Lamantia recently wrote about learning from games. People approach games in different ways—as a personal escape, a way to socialize, a form of mindless entertainment that lets them immerse themselves in a storyline, or a way of challenging their skills against others. A game like Eve Online covers these different styles of game-play well, providing rich storyline content and opportunities for solo play with cooperative opportunities for groups numbering in the hundreds.
A large-scale, multiplayer online game should incorporate design features that combine each of these styles and approaches. Understanding the combinations of features that support different game-play styles can require a design researcher to incorporate observational research with interviews, ethnographic, and sociological research techniques. Such combinations of features tend to be fixed, but the expectations of players within each segment can vary—and it is these variations we need to understand.
Segmentation As Framing
When we adopt a particular audience segmentation model, we also adopt a particular perspective on a problem space that, in turn, shapes our understanding of the solution. This framing of the problem allows us to focus our efforts, but can also leave us blind to other, potentially significant characteristics of a potential solution. For example, if we use market segmentation and focus on the differences in people’s purchasing triggers, we may miss opportunities to provide features that customers have not requested, but would offer significant value. Alternatively, a focus on the lifecycle of the customer experience can hide differences in audience needs within a particular stage.
In the same way that we can overcome shortcomings in one research method by undertaking several, complementary forms of research, combining or using multiple audience segmentation models can help us overcome such structural blind spots, arising out of a particular perspective on a problem.
Conclusion
The use of a variety of different audience segmentation models can inform the design of products and services in different ways. Perhaps just as important, we can apply the underlying research techniques we use for one audience segmentation model to the construction of other models. The key differences between the audience segmentation models I’ve described in this column lie in the goals of the research and the characteristics we want to explore. Thus, we can build on the work we do to produce one type of audience segmentation to develop other segmentation models—and continually expand our toolkits.
I’d like to thank Livia Labate and Janna DeVylder for their help in conceiving and refining the ideas I’ve expressed in this article.
Great stuff as usual, Steve! I definitely think that a hybrid approach like you describe at the end is the most effective way to go, but aren’t personas hybrid by definition? Sure, they focus on psychographic segmentation, but not to the exclusion of other models. What am I missing?
Very nice and clear overview of the salient differences between these models. My take on models is that they perhaps best frame the designer’s approach and, hopefully, that of the client also. And it seems we recognize that the chasm separating the designer/client from the user is one to be narrowed, if not broached.
In social interaction design, we often have the advantage of being deep users of social tools already—so personal experience can supply some of what a model would otherwise offer as description. Here, though, our social and communication competencies differ, our personalities play a more substantial role—social is about other people—and, of course, our personal experiences and histories with tools vary. Some amount of modeling in the social domain can be replaced by referring to common interaction patterns and by reading user contributions for what it appears that users are doing.
I think we have a lot to gain if we move away from models based on needs—Will’s approach is an example of this—and to experience instead. But our need for that varies, of course, with the interactions we expect of users: with a brand, service, or with social, for example. The more deeply interactions are involved with relationships, the more difficult this becomes to model, for social dynamics come into play.
But this is an area ripe for exploration, particularly now, as brands and businesses recognize the value of designing for social engagement.
Thanks for the comment. Regardless of what information is used as an input to personas—as they’re generally used—the end result tends to be a number of user segments differentiated by behavior. Characteristics are layered in—such as frequency of activity, preparedness to purchase, and the demographic ‘spices’—resulting in a richer image of the user, but not in a way that changes the underlying segmentation type.
When I talk about combining segmentation models, I’m thinking about generating more segments, not a richer view of each existing segment. To illustrate that point, imagine you’ve got four pieces of fruit: a normal red apple, a green Granny Smith apple, a Blood orange, and an unripe normal orange, which is green. We could segment these by color—2 red, 2 green—or by type of fruit—2 apples, 2 oranges. In each case, I’ve got 2 segments.
If, instead, I combine those two forms of segmentation—color and type—I get four segments—each more detailed than either of the original pairings. I don’t believe personas, as a form of segmentation, achieve this. We instead get a richer view of a single archetype, but we still don’t have the additional variety in the segments.
Adrian was too modest to say so himself—and was also nice enough not to point out my omission—but he’s done some great work around segmentation based on social competency that you can—and should—see: “Social Media Personality Types” and “User Competencies.”
Are there ways you can bring the team back to focus on the audience model—to refocus on a specific feature or function and ensure people are grounded around user needs?
Seems like it’s a real balancing act between time spent designing, coding, and reviewing user needs—using personas or the other tools mentioned.
I started in usability quite awhile ago, before much was available in the way of Web usability education, especially in my area of interest. I taught myself testing techniques working alone. Coming from a profiling/sales background with lots of research in persuasion, I assumed that all my test candidates needed to be tested for personality profile and learning styles—after the testing was done.
I assumed that I needed to index their frustration levels; test them in the environment where they would likely be surfing; and measure—as best I could—brand impact, short- and long-term messaging, and emotional responses via body language, since we know that words often misrepresent, but the body can’t lie.
So, for over ten years, all my tests have included this data. It wasn’t until too long ago that I realized that few, if any UX testers were doing this. For example, when looking at the test results for Visual Learners compared to Linguistic or Intuitive Learners, there are some surprising differences, which give a great deal more insight into a site’s messaging, than if one doesn’t review that. If you happened to have 3 out of 5 strongly visual learners on a test—most of whom quickly discount text and care greatly about graphics, layout, and non-textual, visual cues—you might infer some very different things about the test results and decide that the text was flawed.
So I respectfully suggest that, in addition to your good suggestions, as part of the segmentation process, it is important to try to determine dominant learning styles and profiles when building personas. Likewise, it helps improve testing both for reliability and quality of insights.
To make it all more real, we are working on a site that we know has a strongly visual audience; so, our site work will be designed and tested appropriately. We need to make sure that our images do message effectively or our audience will bolt.
For more public-facing sites, we test to make sure that multiple learning styles are accommodated where possible.
Please feel free to contact me if anyone has more interest in this, I would love to expand the database of information on this. I have found it to be a powerful tool in the persuasive site design arsenal.
This is a really interesting perspective and, as you say, something we can apply equally well to design as evaluation. Can you share a reference or two for people looking to learn more?
I don’t really know of anyone in the UX world who’s using learning styles besides me. I am sure they are out there. Another approach that some use is to develop personality profiles either by observation or using the Myers Briggs database of profiles by business category. There is some good messaging information there I am told, but I use our in-house profiling tool instead.
I am not being very helpful here, I realize, sorry, there isn’t much out there. I certainly hope to spur more interest in it though, because it has proven to be very helpful. Please feel free to contact me for more information.
One place to start is to read up on learning styles—Multiple Intelligence—then tests for how to identify them. With that in hand and some experience in persona work, it seems to fall into place readily. Still, it is not part of the typical curriculum for UX designers—though it makes sense to me—so I don’t think it will go mainstream soon. I am easy to find on the Web for anyone who wants to discuss this further and happy to share.
When I read the part about mental models, it also reminded me of the Task-Based Segmentation that Indi Young recommends as part of the process of creating mental models.
I’ve become a fan, because you can pretty much pull it out of your ass in an hour of brainstorming and spreadsheet jockeying, knowing that it will be validated with the research that follows it.
It’s used to help you identify who you want to interview as part of your data collection to inform the mental model, but I think it’s an approach that is good in its own right that can be useful even if you don’t get to the level of sophistication of a mental model that is comprehensive in mapping out all the user tasks and mental spaces.
I think, in an environment where you have a lot of existing research—that is probably not very well utilized—or a team that has a lot of audience knowledge�that isn’t very well documented—this is a good exercise that can produce pretty good images of who the audience is based on existing knowledge. (Though, yeah, the benefit of doing it to inform research is that you can just make it up and validate it through said research.)
I recommend it for its simplicity and the value of the output. Here’s an example
of something you could end up with. I personally have further bastardized the approach and made it even more useful to me in a project where I had so much existing research on a complex topic that it was hard to make sense of the audience needs and expectations. By getting to the point in the exercise where you identify a continuum to express each distinguishing attribute across the audience—like so—I was able to represent really important behavioral patterns that were present in every piece of our research we had, but had not been articulate very efficiently in any of them. Big win for me and all the future work we did and are still doing in that space.
Focusing on the business side of the user experience equation, Steve has over 14 years of experience as a UX design and strategy practitioner, working on Web sites and Web applications. Leading teams of user experience designers, information architects, interaction designers, and usability specialists, Steve integrates user and business imperatives into balanced user experience strategies for corporate, not-for-profit, and government clients. He holds Masters degrees in Electronic Commerce and Business Administration from Australia’s Macquarie University (MGSM), and a strong focus on defining and meeting business objectives carries through all his work. He also holds a Bachelor of Science in Applied Statistics, which provides a strong analytical foundation that he further developed through his studies in archaeology. Steve is VP of the Interaction Design Association (IxDA), a member of IA Institute and UPA, founder of the UX Book Club initiative, Co-Chair of of UX Australia, and an editor and contributor for Johnny Holland. Read More
is also used when a company divides consumers with identical product or service needs into groups, so they can charge them different amounts of money.
