What is user segmentation and why is it important?

User segmentation is an incredible tool to find insights that can lead to new or better features.

When you do user segmentation well, you mitigate the risk of missing insights based on the average.

When you just consider the aggregate and the average, you miss critical insights on ways to improve or worse: assume something is going well when it isn’t.

One of the best tools to begin learning about users is user segmentation. Segmentation helps you identify what characteristics imply real differentiation in the user base. Furthermore, it allows you to get so much more value out of additional analysis you might layer on.

“The process of defining segments is a search for potential opportunities: which scenarios, users, or features are successful in my product? Which ones aren't? Great product managers decide on the right feature or target market by conducting a sufficient opportunity search via segmentation.”



— Crystal Widjaja, Chief Product Officer at Kumu

In this post, Shaun Clowes and Crystal Widjaja will dive into:

• What is user segmentation and why is it important
• The five most common segmentation models
• Three steps to a user segmentation analysis

Meet the Contributors

What Is User Segmentation & Why Is It Important

Conceptually, user segmentation is the process of dividing up users and grouping them. But more specifically, user segmentation is looking at how a specific population contributes to a metric over a period of time. It is the key that opens up the door to driving true value for your users and your business.

There are three main ways that user segmentation analysis drives value:

1. Drive changes in user behavior by identifying and focusing on gaps
2. Acquire more of a certain segment by understanding who is successful
3. Create distinct experiences and outcomes for different segments

For example, consider a scenario where you have segmented your users by country, and in each country, you are looking at the acquisition rate over time. In this scenario, we can look at Spain, Germany, and France to see there are differences in the acquisition rates and the trends over time.

Even with such a basic user segmentation analysis, this can help you decide the best ways to move forward. If you are looking to drive user behavior changes, you could focus on the fact that acquisition in France is far lower than in Spain and Germany.

Thus, you’ll want to drive users to activate more specifically in France, and might do that through improving French translations or communications in the activation flow.

If you are looking to acquire more customers that could be successful, you can see that geography, in this case, makes a difference on whether or not someone successfully activates.

A company might choose to focus on those who are doing well and grow that pool instead of trying to change those who are not yet successful on the product. In this case, they would then invest resources into acquiring more users in Spain and Germany.

User segmentation can also help you create tailored experiences. Your users could be identifying extremely different value propositions, and you could use segmentation to serve each of those rather than trying to serve all users in the same way. Or, you could have very different monetization approaches based on what users you are targeting. Without segmentation, you don’t have those levers to pull.

Five Common Segmentation Models

In theory, there are an endless number of characteristics with which you could segment your users. However, you cannot segment users on just any characteristic; you need to have data that enables the identification of if a user has a particular characteristic. As a result, Shaun and Crystal find there are five segmentation models that are the most common.

Segmentation Model 1: Demographics

The first segmentation model is likely the most intuitive: Demographics. Demographic segmentation uses demographic attributes, such as gender, language, race, and geography to divide up the user population.

This is useful when our users derive meaningfully different benefits from the product, or when they use the product in notably different ways.

Demographic segmentation is of limited use because it can be very difficult to get the data. In many cases, demographics must be inferred, rather than completely known. This is especially true for B2B products, where demographic information is relatively limited or inactionable.

Segmentation Model 2: Device Type

The second segmentation model teams can leverage is device type. Device type segmentation analyzes the differences across platforms, including desktop, mobile, and tablet.

Device type segmentation is particularly useful for answering certain questions, such as:

• Do desktop users versus mobile users behave differently across device types?
• Are updates to your apps improving the experiences of the different groups of users? Or are updates introducing problems that you haven’t yet identified?
• Are there features or capabilities that you haven’t enabled on certain platforms that might meaningfully impact overall user retention?

Pro tip: If you are going to segment by device, it is important to segment by app release version. If you don’t, you’ll miss trends across releases, because not all users use the same version.

Segmentation Model 3: Acquisition source

The third segmentation model is acquisition source, which is used to analyze differences in users based on where they were acquired. Common sources include organic, google search, and facebook ads.

Acquisition source segmentation is useful for learning more about where users are coming from, as well as how that influences their behavior on the platform. It’s also helpful to determine the effectiveness of growth efforts to acquire users.

However, teams cannot often influence acquisition sources, so while it can be a very useful starting point, it’s important not to get stuck constantly examining this.

Segmentation Model 4: Persona

The next segmentation model, Persona, uses predetermined product personas to differentiate between users.

This segmentation model is common because the data is often easy to obtain through self-reported surveys. It can help us understand questions like: Who is my target market? And do different areas of my target market perform actions that are dramatically different from others?

However, it’s important to understand that this type of segmentation is of very limited value when the personas themselves are built based on self-reported data. This is because self-reported surveys offer no incentive for the survey takers to tell the truth, often resulting in faulty data.

Segmentation Model 5: Product Category

The last common segmentation model is product category. Product category segmentation is done using company-determined user categories within the product, often payment or engagement tiers.

Three Steps to User Segmentation Analysis

Now that we’ve discussed what user segmentation is and the different basic types of user segmentation, we’ll dive into how to actually perform segmentation analysis.

There are three steps to completing a user segmentation analysis.

1. Define your variables
2. Pull and organize the data
3. Analyze and draw conclusions

The most important thing to remember throughout the process is that segmentation analysis isn’t helpful unless it is actionable, so you want to always try to make decisions that can lead to action rather than just interesting tidbits.

Define Your Variables

As you know, user segmentation is looking at how a specific population contributes to a metric over a period of time. Variables of segmentation analysis break down neatly into four elements: segmentation model, inclusion criteria, behavior, and time period.

First, you choose a segmentation model. Focus on picking a model that you believe would impact user behavior for the business.

Pinterest, for example, might believe there are meaningful differences in how gender identification impacts behavior, and would choose a demographic segmentation model. They might look at male-identifying, female-identifying, and non-binary users.

Second, identify the inclusion criteria: what a user needs to do, and when, to be included in your analysis. This inclusion criteria will depend on why you are doing the segmentation analysis. For our Pinterest example, this could be all users that have completed a sign up in the last year.

Then, pick the behavior you are measuring. This is an action that often connects to an important business metric. Pinning and repinning is a strong engagement metric for Pinterest, so they might choose that as their behavior.

Last, determine what period of time you are going to look at. You often want to consider looking at data over multiple months if possible, so you can avoid any noise coming from seasonality, promotion periods, or even just initial positive results that do not sustain over time. With Pinterest, we will consider looking at up to 9 months since sign up.

Knowing what datasets you have access and can use is an important part to defining the variables. As we’ve said before, you can’t analyze something you don’t have data for! With your analysis now defined, you can actually begin to do it.

Pull & Organize The Data

The next step is to pull the data and organize it in a way that shows potential insights.

Most segmentation analyses do not require complex data, so usually product managers could pull what they need from whatever self-serve tool your company uses. If the data is not all in the same dataset, you might need to adjust it to become useful.

In the Pinterest example, you want to pull the number of pins and repins per user that has signed up in the past year. To be useful to your analysis, the data needs to be organized into a table with three rows: female-identifying, male-identifying, and non-binary.

Each column should represent a month after the sign up date, and the intersection between gender identity and month should be the average number of pins and repins. Once you have this table, you can visualize the data in a line chart.

The first thing the line chart does is provide you with an indication on whether or not the segmentation analysis will be useful. You’re looking for a chart where the lines, or segments, are meaningfully distinct from one another and move in a predictable fashion.

The lines moving in a predictable, different way is what lets you know you’re using science to decide what to pursue rather than guesswork.



—Shaun Clowes, Chief Product Officer at Confluent

The next thing the line chart does is help you draw conclusions from the data.

Analyze & Draw Conclusions

There are two parts to drawing conclusions well: the first, is to understand what that data implies. The second is to determine if there is value or action in that implication.

In this Pinterest example, we can see that users who identify as female tend to be more active at pinning than those who identify as male. While that appears to be true, you need to take this a step further for a full segmentation analysis. Ask yourself: with this information, what might your company do?

One option for Pinterest could be to try to customize the user experience for each segment based on their already existing behavior. This could mean providing more pin and repin experiences to female-identifying users, while focusing on a different set of engagement features for male-identifying users.

User Segmentation Identifies New Opportunities

Learning how to segment users, source insights from the data, and put together an action plan is an invaluable skill for product managers.

When you segment the data, you can figure out who is successful and hypothesize around why, create unique experiences for targeted parts of the population, and identify and focus on gaps in how people use your product.

With the right analysis and bias for action, user segmentation becomes a tool that you can leverage to level up your product in ways you wouldn’t have expected!

For a deeper understanding on how to use data, consider checking out Data for Product Managers on Reforge today.

Sign Up Now