SHARE THE ARTICLE ON
A cohort is a group of users who have a common trait across a specific time period. Cohort analysis is the examination of these users’ shared features across time.
A cohort is a group of users who have a common trait. Cohort analysis examines those users’ retention data over time. There are two primary methods for categorizing our users for cohort analysis:
Acquisition cohorts: categorize users based on when they first joined up for our product. With a consumer app, we could segment our cohorts based on the day they joined up. Monthly cohorts are more likely to be tracked by SaaS providers (such as Amplitude). We can assess how long people continue to use our app from their initial start point in these cohorts by evaluating retention.
Behavioral cohorts: split users based on the behaviors they display in our app during a specific time period. These might be any number of distinct acts that a user can take, such as uploading a photo, listening to music, purchasing gold coins, or any combination of these actions. A cohort is a set of users who completed those activities during the timeframe we specified (for example, within the first 3 days of app use). We may then examine how long various cohorts remain engaged in our app after completing those steps.
Because DAU / MAU numbers are highly skewed by growth, cohort analysis is crucial. If your app is fast expanding, new user signups will disguise where your old users are leaving in your DAU /MAU numbers.
Cohort analysis is a more effective technique of looking at data. Its use is not restricted to a particular industry or function. eCommerce organizations, for example, can utilize cohort analysis to identify goods with the greatest potential for sales growth. It can assist discover web pages that perform effectively in digital marketing based on time spent on websites, conversions, or sign-ups. This study may be used in product marketing to determine the success of feature adoption rates as well as to minimize churn rates.
Cohort analysis is commonly utilized in the following industries:
Cohort analysis is routinely used in all of these sectors to understand why consumers depart and what can be done to keep them from leaving. This gets us to the Customer Retention Rate calculation (CRR).
This formula is used to compute the customer retention rate.
((E-N)/S) X 100 = CRR
The formula is made up of three parts:
E – At the conclusion of the time period, the number of clients.
N – The number of new clients gained within that time period.
S – The number of clients at the start (or commencement) of the term.
To calculate customer retention, subtract the number of customers gained during the time from the number of customers left at the conclusion of the term. This provides an accurate image of retained consumers. To calculate the proportion of consumers who have remained loyal since the beginning, divide the total number of customers by the total number of customers at the start. This is the rate of client retention.
A higher CRR indicates more client loyalty. You may determine where you rank in terms of client retention by comparing your company’s CRR to the industry average. If CRR paints a grim image, remedial steps may be implemented via data analysis – this is where cohort analysis comes in.
Conducting exploratory research seems tricky but an effective guide can help.
Let’s pretend we have a music app. We’re aware of a churn issue, but that’s about it. We’re keeping an eye on the figures, and we’re losing roughly 10% of our users per day.
The users in the above retention data are divided into daily cohorts – those who joined up on the same day. On October 27, 13,473 people joined our music app. Day one retention was 57.9 percent, day seven retention was 19.9 percent, and day ten retention was 14.1%. So, one in every five users who signed up on October 27 were still active users of the app on the seventh day after initially using it.
Out of all of your new users over this time period (322,902 people), 49.8 percent are retained on day one, 15.8 percent on day seven, and 4.7 percent on day thirty.
The easiest approach to view this data is to create a retention curve that demonstrates your retention for various cohorts over time. When you chart your data in this manner, it is quite simple to determine when consumers are abandoning your product.
This retention curve tells you something very crucial right away: 50% of all users abandon the app after the first day.
After the initial huge decline, there follows a second sharp decrease to less than 20%, before the curve begins to level off after a week, leaving roughly 5% of the original users still active in the app at day 30.
If our app exhibits the retention curve illustrated above, we’ll want to know what we can do to improve retention.
The issue is that simply looking at acquisition cohorts provides little insight into how we can enhance the user experience to retain our users. It is not possible to isolate certain actions or user attributes.
Acquisition cohorts are wonderful for identifying patterns and alerting us when individuals are leaving, but to understand why they are leaving, we must turn to a different form of cohort: behavioral cohorts.
Users make hundreds of decisions and demonstrate numerous little behaviors from the minute they sign up for our app, all of which contribute to their decision to remain or go. These behaviors might be anything:
utilizing key feature X but not utilizing core feature Y
engaging solely with alerts of type Z
connecting with 1-2 persons on the app rather than 10+
That implies that if we decide to redesign our user onboarding to increase D1 and D2 retention, we’ll have a million suggestions for how to achieve it. Finally, we may go for the option that was most strongly fought for by the product team’s loudest member
We may build distinct user cohorts for our music app by doing the following actions: playing a song, searching for an artist, or making a playlist.
Assume we wanted to examine the retention rate for app users who favorited music. We may utilize behavioral cohorts to look at retention among new users who favorited three or more songs:
While 50% of all users abandon the app during the first day, just 15% of users who like three or more songs abandon the app after the first day. Favorite songs may be something that keeps people coming back.
To double-check, we may look at the opposite — all users who did not like three or more songs:
Users who did not like three or more songs have worse retention than the average, with 55 percent leaving after the first day.
Even from this modest research, it is clear that allowing consumers to like songs early in their experience helps them to encounter the app’s primary value, implying that they are more likely to stay as users.
Of course, a music app is more than just a place to save our favorite tunes. Behavioral cohorts provide a more detailed knowledge of why consumers leave.
To generate behavioral cohorts, we may utilize any action available in our app. This implies that any number of distinct user activities may be linked to retention rates.
Our music app, for example, offers a wonderful feature that allows individuals to join groups based on their preferred genres. So we’re wondering if it helps with retention or if it’s just clutter that confuses the user.
We can see that early retention is similar to users who love songs, but it’s also clear that retention is greater for those participating in these groups in the medium term, i.e. beyond the first few days of using the app.
As users become more connected with other people and discover new music to listen to, they begin to like the app more and continue to use it. In your retention process, we would most likely hypothesize this, but now we have statistics to back up our theory in black and white (well, burgundy and mint green).
The next step is evident from here. What about users who like music and participate in communities?
As we can see, users who demonstrate both of these behaviors are considerably more likely to continue using the app in the first several weeks. At the conclusion of the first week, the favorite+community cohort has a retention rate of more than 60%, while those without either of these activities have a retention rate of less than 10%.
The most significant element of cohort analysis is that we will not only observe who leaves and when they leave, but we will also begin to understand why our users quit our app—allowing us to repair it. The procedure can be divided into many stages:
Goals – Establish a process goal. Is it our goal to decrease churn in the short term? In the long run? What is our projected growth rate?
Exploration – Investigate existing data to see where modifications may be made to achieve our aim.
Hypothesize – Determine which questions to ask and the potential results of studies.
Brainstorming entails imagining possible experiments to test hypotheses.
Testing entails carrying out various tests in order to assess theories.
Analyzing – Examine test results to see whether or not objectives were reached.
Systematizing entails incorporating any good modifications into the system.
You’re looking at a certain stage in the user lifecycle at each step, and then applying different techniques to improve retention in each:
How do the retention rates look on day one (D1)? Do you witness a decline to 80 percent active or a drop to 10% active? How soon can you lead your consumer to that “a-ha” moment when they grasp the true worth of your product?
Mid: from D2 to D30, are you able to continuously bring the customer back and let them experience the app’s primary value on a regular basis?
Long-term retention often plateaus and becomes asymptotic. The greatest apps feature a retention curve that resembles a grin, with user reactivation resulting in an upward slope later on.