Dive into the dynamics of the client base: cohort analysis and flow analysis

I continue the series of articles on product analysis ( start )

In the last article, I plunged into the analysis of revenue and broke it into 2 components - MRPU and the number of clients. Today we will consider the next steps in the analysis and decompose into components the number of clients and their dynamics.

Now the general analysis scheme looks like this:

A cohort analysis allows us to explain the trends in the client base and sends a direct bridge to the sales funnel and actions to retain and return customers.

What is a cohort analysis? This expansion of customers on the dates of their "arrival". For different products, this may be different events, for example:

  1. First purchase
  2. Signed Subscription Contract
  3. Decorated paid service in the downloaded application.
  4. The first transfer of money to the personal account

It all depends in the end on your definition, when you consider that you have a client. It is more logical to be attached to the moment of earning an income or when a client has an obligation to pay something. Although each product may have its own characteristics and the client can already be considered and the one who signed a contract, even without making money.

If we divide all customers by “arrival” dates, group them by months (or weeks, days depends on typical life cycles of customers) and calculate the number of customers who still continue to be a customer (still paying, not terminated the contract) we will get Something like this:

For ease of analysis, cohorts that are often close in date are combined so that the diagram does not look like noodles.

In my example of attracting clients, everything is fine and the client base is growing by attracting new clients. At the same time, at some point it is possible to return old customers (we see that the oldest cohort increases by the end of the period).

In the analysis of cohorts, we have a number of important derivatives of characteristics that are worth paying attention to:

  1. The size of the new cohorts is a direct characteristic of your efforts to attract customers. New cohorts are formed from new customers.
  2. The cohort decay rate is the average value with which your new customers decrease over time as their lifespan increases. This is usually the percentage by which the cohort decreases over a period of life.
  3. The dimensions of the "old" cohort. In the "old" cohort usually place customers that you no longer find new. These are people who should in theory be your regular customers. Most often, this cohort forms the bulk of revenue and the largest in terms of numbers. The dynamics of the size of the "old cohort" determines your prospects as a product. Reducing the "old cohort" or its stagnation - a bell that you have a problem with the product, with sales or loyalty.

I want to note that usually there is no "life" of the client, because most often, cohorts last and last, they simply have less and less clients. In this sense, the everyday meaning of the word "term of life" turns out to be an incorrect interpretation of the decay of cohorts. If we are talking about 3 months of “average life”, then it is not true to understand that you have no customers left after 3 months. The use of the term "average life" becomes a mathematical trick. The fact is that the decay of a cohort is characterized by the rate of decrease of customers. And you can translate this tempo into terms: I lose 50% of the cohort in 3 months. Or even harder - I lose 95% of the cohort in 12 months. But it is possible that the typical period of loss of the entire cohort will stretch over a year. Therefore, it is good to clarify in your analytics,

Using the metric "X% for Y periods" is a good quantitative method for comparing the quality of cohorts among themselves. The fact is that any cohort is a small "experiment". People in each cohort are meeting and getting to know your product from scratch. And a historical retrospective of a cohort shows your successes and failures with respect to onboarding, and then with respect to retention \ churn. If you are methodically seeking to improve the characteristics of Y, then this means that you are well developing the product and customer relationships. In general, this is a matter of taste, you can both operate on the "average lifetime" in the sense in which I outlined it above, or use the cohort breakdown percentage.

Another good method of “looking” at a cohort is the analysis of customer base flows. This is a more convincing convolution of data in cohorts. We connect our cohorts and their dynamics as follows:

  1. How many new customers came in the reporting period (just new cohorts)
  2. How many old customers returned in the reporting period (customers from the old cohorts, then renewed the relationship)
  3. How many clients are in the operating base (were and still are)
  4. How many customers left the old cohorts

This picture clearly shows the balance of the inflow and outflow of customer base. And if your outflow is higher than the inflow, you immediately understand that you have problems.
In this example, the balance of customer flows is strongly in the direction of inflows and therefore the customer base is growing rapidly.

What gives us the decomposition of clients into cohorts:

  1. We can see how quickly our customer base is being updated, which part of it consists of "newbies", and which part of "oldies"
  2. If the basis of the customer base is old people and your new customers give 1% to the base per month, then it is somehow strange to expect a customer growth of 50% by the end of the year. You either have to increase the incoming customer flow (which is usually easier) or return a significant portion of previously lost customers (which is usually more difficult).
  3. If the life of your customers is short and you don’t accumulate "old men" almost at all, on the contrary it means that your efforts in relation to old people should be in the background. And you need to work on increasing the life of the client, onboaring or continuing to increase the incoming flow.
  4. Cohort analysis gives you the opportunity to predict the future state of your economy and answer the question "Can you fulfill the plan, if at the same decay rates, you double your customer acquisition?"
  5. We can quantify the success of our efforts to attract and retain customers through comparing the decay rate of cohorts.
  6. According to the results of the analysis, you can understand where the problem is in the product in relation to the customer base (attraction, retention, “life”, etc.)

I have already noticed that while we are on the analysis of all averages. But your customer base may have different segments and clusters. Your cohort analysis will be even more productive if you carry out the segmentation and clustering of the client base.

In the next couple of articles I will write on preparing data for cohort analysis and on predicting cohorts into the future for predicting the dynamics of the client base.

Also popular now: