Data-driven companies understand that the individuals who rely on their products are not a monolith. Users find their way to a product by way of vastly varied paths and with their own unique objectives. Given this, it is crucial to make decisions with the wide-ranging needs and wants of those users in mind.
User segmentation is the process of grouping your users based on similar characteristics or patterns of behavior. Learn more about what user segmentation looks like with these examples from best-in-class SaaS companies.
The reasons to segment users crop up across all aspects of a business, but the best reason for embracing user (or customer) segmentation is simple: segmenting your users is ultimately a reliable way to deliver more value.
Creating a set of meaningful and clearly defined segments is a crucial step on the path to becoming more data-driven. Check out our segmentation guide and worksheet for step-by-step instructions on how to get started with segmenting your users.
Top 8 user segmentation tips
User segmentation isn’t an exact science. Still, there are some useful guidelines to keep in mind as you begin the process of segmentation. Below is a short, surely not exhaustive, list of tips for that process:
1. Begin segmentation with behavior (then characteristics)
This is in part because user characteristics are often more explicit anyway – like the region a user is in or the industry their company might be a part of – and available due to self-reporting. More fundamentally though, understanding user behavior is generally more critical to improving a product than understanding characteristics.
2. Prioritize segmenting users where there are blind spots in your data
Wherever questions exist about user behavior, that is likely the most helpful place to begin analysis for segmentation. If engagement habits are somewhat unclear for example, work to segment the varying levels of product usage.
3. Clearly articulate the question(s) you’re trying to answer
Write them down! Those questions might extend to particular user behaviors or product features, but in either case having an explicit question or goal is crucial. Often, writing those points down is a good forcing mechanism to hone an analysis of any kind (segmentation included) to your chosen goal.
4. Make sure your segments are MECE (mutually exclusive, collectively exhaustive)
This should be the case for the population of users you’re interested in. If you segment a population of users and have a meaningful number of users who don’t fall into any segment, you’re missing some of the story.
5. Limit the segments based on any particular trait/behavior to a reasonable number
Anywhere from 3-8 classes is generally best for being both (I) helpful/descriptive and (II) not overly-specific. For example, segmenting user engagement as simply either engaged or unengaged isn’t terribly informative; meanwhile, creating 40 ranked and defined levels of engagement can be confusing and less actionable. A constructive heuristic might be: if you have a line graph grouped by your user segments and you have to squint at it to see what’s happening, you have too many segments.
6. Understand the type and distribution of your data as your segment
Questions one might ask themselves include: are you segmenting numerical or categorical data? Is that categorical data ordinal? Are users distributed roughly evenly across categories or are some categories very large/small? Is numerical data roughly normally, uniformly, or otherwise distributed? Understanding these questions helps teams later define more useful segments.
7. Check if your segments behave differently across metrics of interest consistently over time
If they do not, either (I) congratulations, you’ve found something particularly interesting! or (II) there isn’t actually all that much of a difference between certain segments (they behave similarly), in which case you can consolidate groups to help you stay within the ~3-8 range above.
8. Seek out tipping points in your data when defining segments
A team might ask itself what separates “moderately” and “highly“ engaged users. The point of distinction ideally is not arbitrary, but defined after understanding the points at which users are getting disproportionately more value.
As a simple example, if using a product 3 days versus 2 days a week doesn’t correspond to a clear jump in value created for a user (evidenced by sharing findings, using the product for additional insights, etc.), but using a product 4 days versus 3 corresponds to a spike in similar indicators for value, highly engaged users might best be defined as those who use a product 4 days a week, as that is the threshold product teams should emphasize.