Imagine you’re working at a startup that developed a new fitness app. Since the launch, you’ve been acquiring users steadily, but you want to understand your retention better. For example, you want to analyze why users who signed up in March didn’t stick around in April, or why those who joined in July are still actively using the app.

It’s a powerful tool that helps you track and understand the behavior of consumers over time. Using this, you look at each group of customers and see how their actions and loyalty change. With this information at your disposal, you can understand how and whom to onboard, solve key marketing challenges and the likes. It may not be surprising how this can affect and assist your growth and product operations.

Therefore, if you're new to the world of Cohort Analysis or want to know the ins-and-outs of it, this guide is here to demystify it for you! Filled with both textbook definitions and comprehensive examples, this will be your go-to primer on this subject. After completing this (and your workout), you will be able to:

Creating a Cohort Chart and being equipped with foolproof analysis so you are able to solve for retention, acquisition, and multiple other problems!

Introduction to Cohort Analysis

Brief introduction to the Foundations of Cohort Analysis and what to expect

Therefore, if you're new to the world of Cohort Analysis or want to know the ins-and-outs of it, this guide is here to demystify it for you! After completing this (and your workout), you will be able to:

Learning is tough. It takes effort. Lots of it. It's not easy to pull the brakes on your work or life and sit down to learn when you have the option to spend your time in a million other ways. 

But now that you've taken this step for yourself, let's do it right. 

Don't think of this guide as just another piece of content to read and learn from. Instead, take it as a step-by-step process, which if you follow will help you achieve the above outcomes. If you find yourself dozing off - take a break, stretch and come back. Take notes, read, re-read, process, practice and apply. 

Set up your work desk and sit there (I know, the temptation to be on your bed is real)

Put your phone on silent, switch it off for the next hour if possible. 

Grab your favourite cup of tea or coffee and settle in. 

Intro to Cohort Analysis

Cohorts and Their Analysis

Understand what Cohort Analysis is and why you might need it.

A cohort is a group of customers who share a common experience within a specific time frame. Let's open your fitness application again. Then, say, all the customers who subscribed to it in January form the January cohort. Similarly, customers who subscribed shop in February are part of the February cohort, and so on.

It’s like detective work for your business! You’ll track groups or “cohorts” of customers over the next few months to see how often they return, what they interact with, and how their behavior changes over time. Whenever you see a metric which isn't moving, you'll be relying heavily on this tool to dive deeper. You’re looking at all the things they do, even if those actions happen months after they are on-boarded. 

For example, with cohort analysis applied in your fitness app example, you can understand how regular your users are with each workout, maybe also how you can correlate their diets to their regularity and so on. Say you're analyzing user behavior relative to the third month after they join. For instance, users who joined in January (M0) would be analyzed through March (M3). You might notice that the January cohort had a significant drop-off by March. This could prompt you to investigate further. Maybe there was a change in your UX that month, or perhaps you increased your notifications being pushed to them. On the other hand, if your November cohort remains loyal, you might want to look at what special promotions or events you ran in November that encouraged them to keep coming back.

You might not always have to run into a problem

 to begin analyzing based on cohorts. Cohort analysis essentially gives you the insights you need to make informed decisions. It helps you refine your marketing strategies, improve customer satisfaction, and ultimately scale your business.

We know. Even before you dive into how you can perform Cohort Analysis, you want to know what's in it for you. Check out a couple of use cases you might often find yourself in!

With a thorough understanding of what cohorts and their analysis are, it's now time to dig deeper. In the next chapter, we will read about:

In classic GrowthX style, here's a whimsical for what the next chapter is going to look like!

What are Cohorts and Why You Need Them

The what's and why's of Cohort Analysis. Understand what exactly we mean by Cohorts and why analyzing them helps you.

How to Perform Cohort Analysis

Understand segmentation and a step-by-step guide to perform Cohort Analysis!

When we started learning about cohorts, we talked about how we can segment them on the basis of any (and every) thing that might be beneficial to solve a particular problem in hand. This could include, like we said, when your cohort graduated, their social media usage, the brand of their phones, and so on. Broadly speaking, there exist three umbrella categories of such common characteristics you can base your cohorts on:

Think of these as user groups based on when they took any specific action.

did they log their first workout in March?

→ Did they make their first purchase in February?

Analyzing these cohorts reveals that consumer behavior is dependent on when they begin utilizing a company’s products or services. 

Two of the key issues these solve are engagement and retention.

For instance, if 70% of customers who signed up with the firm in the first quarter stay with the company in the fourth quarter, but only 30% of consumers who signed up in the second quarter stay with the company until the fourth quarter, it indicates that Q2 customers were dissatisfied. The corporation may have overpromised during Q2 marketing, or a competitor may be offering better products or services to the same customers.

This is where things get interesting. Here, you group

, not when. Think habitual users of your product/service, regular users of other products in your category, or folks who love to explore beyond their absolute needs. 

Understanding these behavioral patterns helps you tailor your approach to different user segments.

 For example, you might be working on boosting sales for an e-commerce company. You can then make cohorts on the basis of your users’:

Upon learning more about all cohorts on the basis of these categories, you might be able to tweak your marketing strategy to tailor to, say, increasing incentives for your Indian market's electronics and furniture customers.

Size-based cohorts group users by the magnitude of their interactions or the scale of their organizations. This type of segmentation helps you understand how transaction size, company size, or usage volume correlates with 

For example, small and startup businesses typically experience higher attrition rates in a SaaS business model than enterprise-level companies. Small and fledgling enterprises may have a limited budget and are evaluating low-cost products to see what works best for them. Enterprise-level organizations have a higher budget and are more likely to stick with one product for a longer period of time.

identify which customer segments contribute most significantly to your revenue,

 which ones have higher churn rates, and how different usage patterns impact customer satisfaction.

Try to identify the type of segmentation we used here (answer in the next callout!):

The goal is to understand and cater to the unique needs and behaviors of different customer groups, which ultimately helps in growing and retaining the customer base.

Phew! That was a lot, wasn't it? Take a break before we read more about how to perform Cohort Analysis. Next up:

How to Segment

We keep mentioning how it’s important to track metrics and see how one cohort is performing when pitted against another. You need the right data to build the right cohort. Choosing metrics also comes with its own science, though. As we’re closer to building our own chart, let’s look at what inputs we need to gather.

This is when you decide how often you will be gathering/analyzing data for your product. Your metrics can simply be classified into leading and lagging indicators. Leading indicators help predict future performance, while lagging indicators provide insight into past outcomes. Tracking both types of indicators will help you form comprehensive understanding of your business's health.

For example, high user engagement—a leading indicator—suggests that customers are likely to stay and continue using your product, which can lead to future revenue growth. Because leading indicators help guide your daily actions and can change quickly, they should be tracked frequently—sometimes even on an hourly basis for dynamic environments like e-commerce platforms (e.g., Flipkart).

Lagging indicators, like plan renewals or monthly recurring revenue (MRR), measure past performance. These indicators are typically reviewed less frequently, such as monthly or quarterly, because they reflect outcomes that have already occurred. For example, in a subscription-based product, you would track plan renewals monthly to assess how many customers continued their subscriptions from the previous month.

There may be a certain set of 10% of your users who contribute to, say, 30% of your revenue. These are who your Power Users are. Our experts have called out how it is important to constantly track the behavior of these users and if it is getting affected in any way.

Different companies have different definitions for the power users. Some experts in fin-tech, for instance, have told us they define power users as not just those who make high-volume transactions but also those who re-subscribe to their product 2-3 times. give paytm example

As you will see later, if you observe a dip/uptick in their behavior, it is time to work backwards and understand nuanced customer behavior for that particular metric. Some products go on to A/B test new features with only 1-2 groups of power users, while others ensure to conduct hygiene check call with them very frequently. 

It is essential to understand metrics which are not working well with your power users; and then track how they behave v/s the next,say, 10% and so on. This is a key trick in planning your campaigns and feature releases!

Just because you can track something does not mean you should. Before you undertake this activity, the most important question you need to answer is 

 More often than not, these problems will revolve around your North Star Metric(s).

While crafting marketing or growth strategies too, you should be aligned with your companies goals, ergo the North Star Metric. For most products, this would be revenue.

Other common problems could revolve around:

Now that you’ve decided what problem to work on, we will choose what metrics you will be focusing on. By analyzing the correct metrics, we can gain insights into how well our product or service is performing, where we can improve, and what strategies are effective.

Some standard metrics are the same as your north star metrics, as discussed earlier. These give our analysis a direction as to what problem we want to solve, in the first place.

There are custom metrics, which provide a second degree of depth and help us see where the problem or a disruption in pattern actually lies. These could include:

Fine, it's fairly easy to know what problem you are solving for and which group of users is getting affected the most, but how does one decide which metric will give us the answers? Now that you know what data points you absolutely need to perform Cohort Analysis, you probably might be wondering how to choose your metric?

Choosing your metrics depends heavily on the stage of your start-up. By this we mean, how much data you have and how efficiently can you track it. We have observed that almost all mature products have found ways to efficiently track their data exactly when they need to. In cases like these, you probably already have (the means to formulate) a solid 

! These are not strictly an either-or, you can choose to use a combination of both depending on what you are dealing with. Let's see what we mean by these:

When choosing metrics for cohort analysis, it’s crucial to break down your business growth into core components. We could do this using the growth equation, a basic version of which is:

This formula helps identify functions and dependables which tell us about factors to optimize for growth by examining key elements such as the number of paying customers, the average amount spent per transaction, and the transaction frequency.

To gain deeper insights, segment these metrics further based on relevant factors like acquisition channels, retention rates, and customer behavior. For instance, if focusing on retention, you might analyze demographics such as gender, age, or location to see which groups are more likely to stay engaged.

For acquisition, consider examining marketing channels to determine which ones attract the most high-value customers. By selecting the right metrics and segments, you can uncover specific patterns and opportunities, ultimately driving more targeted and effective growth strategies.

This is one of the first steps you undertake to decide what works for you to solve your problem/standard metric. You need to come up with a hypothesis of an action/behavior or some form of segmentation that could lead to, say, better retention. After coming up with multiple hypotheses, you can compare cohort charts that could lead you to focusing on the right cohort/metric, or act as a tie-breaker in deciding what works.

For example, if you own a social music app and want to increase retention, you could hypothesize on the basis of day 1 actions of a user. You may then make cohorts on the basis of, say, those who finish setting their accounts up and those who don’t.

This method of choosing your metric involves you to work backwards if you are solving for something or just brainstorm if you're ideating on a campaign or feature release. For instance, for a geography-based vernacular app, having data for a country as diverse as India might become an issue for most products. However, when solving/brainstorming for these cohorts, you need to be involved in hypothesizing on every touch-point of user journey and come up with solutions.

You have now understood how to cherry pick your metrics, leverage your power users, and various other inputs. In the last section of this chapter, we finally delve into how you use Mixpanel to generate a Cohort Report and how you read a heat map. We hope you're being able to follow what we have learnt so far!

Building a Cohort Chart: Inputs

Let’s solve for retention using a very popular analytics tools, Mixpanel.

The easiest way to create user cohorts in Mixpanel is through 

. Let’s see how to do it through the Retention report.

 From the top menu, click on Reports and choose Retention.

hat’s where you choose the time frame (

 Save individual cohorts for further analysis.

Most tools work on a similar basis as Mixpanel and might have only slightly differing lexicon. There are several other popularly used tools for Cohort Analysis:

The real meat of any cohort analysis report is the heat map. This heat map allows you to quickly identify the highest (and lowest) performing metrics by cohort. Let’s use this 

C5 (red) represents the number of new customers we acquired in the month of Jan. D5 tells us the number of customers who were acquired in Jan but they returned in Feb.

E5 (orange) is the number of customers acquired in Jan who returned in March.

Basically, as we move along the Jan’s row, we understand how the retention of new customers acquired in Jan fluctuated until Dec.

This table (Table 2 of the sheet provided) gives the exact same data as before, but in a different format. Now for instance, in the fifth row, as we move along, we would know how many customers (in %age) acquired in Jan returned in the succeeding months.

Kudos on learning how to build and read a cohort chart. We are slowly inching towards thoroughly understanding how to analyze the insights we have in abundance. In the next few chapters we see:

Building and Reading Cohort Charts

The faster problems are solved and fixed, the more likely your product will be successful. Cohort Analysis is a part and parcel of Root Cause Analysis (RCA). It provides the essential fodder (read: insights) you need, to be able to identify exactly what is wrong. When you see KPIs change, you look at all the different factors that caused the change.

Any association between a new launch/change in product and a change in metrics is a correlation, not a causation. Hence, you keep drilling down through the cycles and ask intuitive or logical questions, finding the data that answers the question. 

There are multiple ways to do an RCA after you have adequate insights from conducting Cohort Analysis. In our previous section we talked about how you need to maximize on each touch-point to understand exactly why a metric is not working for you. We also learnt through a small example how to perform Change Analysis. In this section, we look at other intuitive ways to do the same, and pin point exactly what led to you cohorts not doing well.

This technique is a simple and effective tool for identifying the root cause of a problem by asking "why" five times. You could find a pain-point in a specific cohort and in that journey, find which metric caused it. Then, you could dive deeper by asking he 

What’s the Problem in hand: Some users are not placing as many orders as they used to

Next up, we look at what could be some problems while performing Cohort Analysis, before we wrap this guide up. You will also see what could possibly go wrong if your company does not perform Cohort Analysis and utilize its insights fully.

RCA using Cohort Insights

Conclusion

Looking back at outcomes we achieved through this guide.

Without cohort analysis, you might rely on overall metrics that don’t reveal underlying issues. For example:

Cohort analysis is like a growth engine to work on any insight and build solutions which are personalized as deep as possible for a set of cohorts, which are not just demographically different.

Before we part ways, we wanted you to make note of one last important thing:

With that we are done with Foundations of Cohort Analysis! It's been a pleasure guiding you through this learning journey, and we trust that the knowledge gained will significantly benefit your processes. Let's look at what we covered in this guide:

We hope this guide has been full of value for you. Consider this a V1 and if you have any recommendations, resources, or information you’d like to contribute, 

Alternatively, you could also email us on learn@growthx.club