How to Use Cohort Analysis to Understand Product Traction

The field of data analytics, within which the cohort analysis resides, is accelerating fast. Statistics are striking:

• The industry’s projection is 30.7% CAGR and reaching $346 billion by 2030. 
• Companies that are data-driven acquire 23 times more customers than those that are not. 
• Moreover, data-driven companies are 19 times more likely to be profitable. 

At the same time, reports indicate that between 60 to 73% of the data that companies collect goes unused. Finally, only 12% of the data that is actually used generates actionable insight. In light of this, cohort analysis is exactly the tool that makes sense of a lot of data and generates actionable insights.  

Cohort analysis is moving to become a staple of product analytics, especially in the SaaS, marketplaces, and mobile apps industries. Historically, it started as more of a B2C analytical tool for tracking consumer behavior to improve retention. One can compare it to studying the behavior of shoppers in a supermarket or mall. For instance, convenience-first shoppers tend to follow predictable paths in the store, have higher basket value, and longer retention over time as compared to discount-first shoppers hunting for marked-down items. Those insights help tailor store layout, promotions, and inventory, among other things. In digital products, instead of tracking foot traffic or checkout lines, companies track signups, feature usage, and repeat visits. When it comes to product traction, different stakeholders understand it in their own set of metrics: 

• Investors will consider Monthly Recurring Revenue (MRR) and other revenue-related metrics, 
• Heads of departments will focus on unit economics and related metrics, and 
• Product people tend to focus on engagement. 

However, at the heart of all these metrics lies the eternal battle for retention against churn. And this is what cohort analysis deals with. 

What is Cohort Analysis

Cohort analysis breaks down users into groups based on a common characteristic and tracks them for retention over time, which can be days, weeks, or months. The canonical form of cohort analysis is time-based, which is shown in the screenshot below from GoogleAnalytics. It divides all users by the time they joined, and then tracks how long they remain in their cohort over time based on a specified condition. Other types include attribute-based and behavior-based analyses. 

Cohort analysis operates with two primary conditions:

• Inclusion criterion – the condition that determines a user’s inclusion into a cohort, which can be acquisition date, user event (sign up, completion of some flow, etc.), or purchase/conversion;
• Return criterion – the condition that determines whether a user stays in a cohort (users who make a purchase, use a certain feature, or simply start a user session the next day, or within a week, or else).

Cohort Analysis Tools

Early implementations of cohort analysis were Excel sheets. Founders, investors, and managers shared Excel templates as practical tools to evaluate product traction before dedicated tools emerged. For instance, the Angel Investor and a Partner at a VC firm, Christoph Janz, said at the time:

“Cohort analyses are also essential if you operate a SaaS business and want to know how you’re doing in terms of churn, customer lifetime and customer lifetime value.”

Here are some of his templates below that include his original commentary.

Today, Google Analytics, Mixpanel, Amplitude, and Tableau are just a few tools in the rich toolkit for performing cohort analysis. However, they vary in their capabilities. For instance, Google Analytics lets you define cohorts based on either time or events, but you cannot break it down further than that. You can specify an inclusion criterion, such as “Users joined between dates X and Y and did event Z”. 

With tools like Mixpanel or Amplitude, you can chain more complex conditions. For instance, users joined between dates X and Y that did action Z 5 times, didn’t do action A, and have a property B. In addition, they support dynamic cohorts, while in Google Analytics, cohorts remain static once created. 

Tableau is a more raw and advanced analytics tool. It does not have a built-in cohort analysis, but if you pre-model data correctly, you can do anything with this tool. 

Cohort Analysis Types

Time-based Analysis

For product traction, the main value of time-based cohort analysis lies in these three use cases:

• identifying feature decay, if your app uses AI/machine learning, 
• testing changes to the app logic, e.g., onboarding, adding new features, tweaking usage caps, etc. 
• linking user retention to the campaign launched at the time of their joining. 

Ideally, when you release updates, cohorts that experience it should have higher retention than the previous ones. Otherwise, it indicates that new changes cause more users to churn. It allows growth teams to ensure that their campaigns set the right user expectations. Also, the product team gets to ensure that onboarding leads new users to immediately experience the value. Finally, the development team can catch model drift for AI/ML functionality and timely repair it.

In time-based cohort analysis, the inclusion criterion is often the time of the first visit or the time of the first session start. In Amplitude/Mixpanel, you’d set this criterion based on the time of ‘Any Active Event”. 

For the return criterion, you would specify anything that equates to your definition of product stickiness. It can be a purchase event within a week, or an event for using some feature within the following day or two. Often, it is simply a session start.

Attribute-based Analysis for Product Traction

The majority of startups use Personas to represent their target audience and drive their MVP development. With attribute-based analysis, you can refine WHO your product’s most loyal users are. The attributes correspond to characteristics like geographic region, acquisition channel, device types, and whatever else is available in your tool. This helps you spend your marketing budget on the audiences that resonate with your product, ensuring higher ROI. 

In terms of the inclusion criterion, it often makes sense to directly link to business outcomes. So you would define something like “Plan Type: Premium” and then track it for different regions. It can be by states, countries, or even more broadly defined areas like EMEA and such. You can often chain attributes, and add to that device types, and whatnot.

For the return criterion, you would simply keep users in cohorts if they log in or have a session within a specified time period. However, a more powerful way is to mix it with behavior-based analysis. In this case, you’d select the return criterion as a value action (success event) that the premium plan corresponds to. This allows you to make sure that users get their value-for-money, and they won’t churn after that paid period expires.

Behavior-based Analysis

This is by far the most powerful cohort analysis type. Diving deep can allow you to mine for the most granular yet impactful insight to drive product traction. In tools like Mixpanel and Amplitude, you can combine it with time-based and attribute-based filtering, as shown in the screenshots below. 

This type of cohort analysis lets you establish a direct link between user actions and long-term retention backed by hard data. The type of insights it produces is:

• “Users who share 3 files within their first 48 hrs are 10x more likely to retain over 6 months”;
• “Users who create at least 2 playlists within the first week are 5x times more likely to convert to a paid plan”;
• “Users who create a shopping list within the first month are 3x times more likely to become repeat loyal customers”, etc.

These insights are critical for product traction. Knowing what makes a retained paying user lets you develop flows to push more users down this path. Behavioral cohort analysis is often central to building viral loops and growth engines. 

For the inclusion criterion, you would specify a particular action for which you want check retention, and frequency of this event per specified time period. 

For the return criterion, you are likely to use the same action or another action that corresponds to user success in your product.

Cohort Analysis for Marketplaces

Product traction in marketplaces depends on the balance between two user groups: buyers and sellers. Sustained growth requires a healthy supply and demand. It ensures that sellers have enough opportunities to transact while buyers are neither overwhelmed nor underwhelmed by choice. 

The basic application of cohort analysis enables studying retention trends for acquisition channels to determine which ones bring the most loyal buyers and sellers. Acquisition is often the costliest item in running marketplaces, as you always have to boost the acquisition of buyers or sellers to ensure a great user experience. 

For this purpose, it is often enough to implement regular time-based cohort analysis. It will help you link acquisition campaigns to cohorts of acquired users and see how they do over time. The screenshot from the Nozzle tool below shows cohort analysis for Amazon’s third-party sellers. It helps them track their ad spend, customer acquisition costs, breakeven month, and retention. 

Other areas where a marketplace can benefit from cohort analysis are:

• Time-based analysis of the churn will help identify when to run personalized re-engagement offers, 
• Attribute-based analysis will show which audience is likely to churn or disengage, so that you might be able to design a loyalty program. Or, you might need to implement UX changes if you see a certain cohort failing to engage using a certain device. 
• Behavior analysis will help design targeted incentives to reach purchasing “cliff” behavior. It might be linked to creating wishlists or completing several purchases over a certain time period, which will be an early signal of future loyalty. 

Cohort Analysis for B2C Apps

For B2C apps, the benefit of cohort analysis for product traction lies in understanding how users discover value and what behavior leads to repeat usage, and – at best – forming habits around your app. Product traction in B2C apps is often fragile. The reality of the niche is low switching costs to a competitor’s app and variable CAC across acquisition channels, which together put pressure on profit margins. As a result, great product traction depends more on user return behavior than on sign-ups. 

The starting use case for cohort analysis in B2C apps is acquisition channels. At a basic level, it helps you discard channels that bring users with short-lived curiosity in favor of channels that produce users with long-term intent. 

Attribute-based cohort analysis helps to reveal structural issues. Often, product traction flattens or drops due to usability issues. It can be localization problems, which will be uncovered through regional cohorts, or device usability issues, which will be shown through cohorts based on device types. Sometimes, it might be a mismatch between real target users and intended user Personas, which can be verified through cohorts based on demographic data. 

Finally, the behavior-based cohort analysis is critical to the success of any B2C app. Insights on what features lead to higher LTV and deeper engagement allow your team to focus on high-impact features and interventions like tailored personalized experiences, timely notifications, and onboarding flows. Product traction increases when the app is built to reinforce ‘success’ behaviors. These can be tracking workouts, logging symptoms, adding meals, and other core actions habitually. 

Cohort Analysis for B2B SaaS Tools

In B2B apps, product traction depends on accounts, how many members of the organization join, in-company adoption, and account expansion. It depends less on individual usage patterns, but rather on time to value and depth of usage.

The foundational use case of cohort analysis in B2B apps is behavioral for completing onboarding and tracking activation. Sometimes, it can be enough to implement time-based analysis, aligning cohorts with changes to onboarding flows. This way, you may not know exactly which onboarding step is responsible, but you will know which onboarding sequence produces the higher number of retained accounts. Combined with behavior cohort analysis, you can track account milestones and link them to particular features. 

Many B2B apps have complex segmentation strategies, and attribute-based cohort analysis is indispensable for this purpose. 

In addition, behavior-based cohort analysis is used for finding early signs of product usage that precede upgrades and upsells. This information is then used to level up the customer education system or even simply use this information for customer success management teams. The latter can be directly used in calls with clients to introduce this type of usage or feature and upsell/upgrade. 

Summary

Type of Cohort Analysis	Use Case for Product Traction
Time-based	Success of marketing campaigns; AI/ML feature decay; Impact of app updates.
Attribute-based	Refining target audience – User Personas; Diagnosing usability issues, such as device-related or localization ones.
Behavior-based	Developing targeted interventions; Driving product development roadmap; Selecting features with the highest impact.

FAQ: How to Use Cohort Analysis to Understand Product Traction