What is cohort analysis and how to do one

Can you imagine having the power to detect exactly when your customers lose interest in your product or service? Cohort analysis allows you to know this. In today's article, we talk about a key technique for evaluating customer retention and optimizing your loyalty strategies.

What cohort analysis is for

We won't get tired of saying it: retaining the customers you already have is much cheaper than acquiring new ones. In this sense, cohort analysis is fundamental to guarantee that long-awaited retention, because it allows identifying what makes your customers stay or leave.

This technique divides your users or customers into groups that share a common characteristic to be able to track how they behave over a certain period of time. This enables patterns that might otherwise remain hidden to be seen and helps to:

1. Know when and why users are lost.
2. Compare acquisition campaigns by real user quality.
3. Visualize improvements after changes applied to the product or marketing strategy.
4. Measure the level of customer acceptance of brand changes.
5. Make better decisions.
6. Identify areas for improvement.
7. Segment the audience better and offer more relevant recommendations.
8. Identify trends within customer behavior.

Each of these sets of users with something in common is a cohort. They are immutable groupings; once formed, there can no longer be new additions or changes. These can be created, for example, by people who registered in the same month, who used a certain functionality for the first time, or who made their first purchase on a specific date, among others. In this sense and, depending on the objective of the analysis, there are two types of cohorts:

• Acquisition cohorts. Users are grouped based on when they were acquired, which can be by week, month, or quarter. They are the most common and serve to detect early churn patterns, evaluate marketing campaigns and how they retain, or make a performance comparison over time.
• Behavioral cohorts. They are grouped based on specific actions such as trying a feature or downloading a resource. That is, it groups users according to activities they do within your application or platform during a period of time.

How to do a cohort analysis

Knowing all the above, now you will wonder how you can do a cohort analysis and start observing your users' behavior. To use this technique, it’s common to use a table. There are many tools that help create them automatically by analyzing different metrics. The most used are Adobe Analytics and Google Analytics.

In these tables each row will represent a cohort and each column is the behavior of that cohort over time. On the other hand, the cells show the percentage or number of users who remain active or who have performed a specific action, depending on the type of cohort you are analyzing. To build a table you need:

1. A reference date to group users (registration, purchase, subscription...)
2. Identify which cohort each user belongs to in order to group them.
3. Define which metric you are going to analyze. Retention, churn, recurring purchases, active sessions, etc.
4. Define the tracking period. It can be daily, weekly, by months, quarters, yearly.

With these data, you can now start building your table and analyzing it. To make it easier for you to understand, we explain it with an example one:

The first column lists the different cohorts; in this case, they’re the months in which the customers were acquired. The second column refers to the users who are part of that cohort. As for the third column, Month 0, it corresponds to the initial moment of acquisition, which is why it will always be 100% of the users; from there, the rest of the months show the percentage that remains active after the respective months.

In the January cohort there are 1,000 users, who were acquired during that month; in Month 1, 92% remained active; in Month 2, 87.5% and so on. The same occurs with the rest of the cohorts.

With this information, what can we observe? In the month of February there’s better retention, since the percentage of users who remain active stays high throughout the months. While in the month of January, retention is much lower. Given this, one would have to investigate the causes behind these data.

For example, February's data could indicate that there was a campaign, a product improvement, or some change that worked correctly and managed to retain users over time.

On the other hand, in January, when worse results were obtained, it’s a sign of some problem that makes users leave and that needs to be solved. For example, it could be a month in which the user experience was worse due to less customer service or failures in the omnichannel strategy, there was some change in the product or service that did not convince users, etc.

Now that you know what a cohort analysis is and how to do it, do you dare to use it in your company?