How to understand and compute your critical cohort metrics

cohort analysis -001.png

Welcome to the second leg of our short article series on cohort analysis. 

In the introductory article of this series, I have presented to you three primary use cases of cohort analysis and key challenges associated with each of those benefits. 

In this article, we will start going over those use cases, starting with the most simple one - cohort metrics. 

We will start this series by first briefly going through the overarching process that goes into the creation of cohort metrics (you can find more details in a longer article published here). 

https://medium.com/analytics-for-humans/a-beginners-guide-to-cohort-analysis-the-most-actionable-and-underrated-report-on-google-c0797d826bf4

We will then go through a few essential cohort metrics for your company and focus on creating a useful framework for you to take actions based on those metrics. 

Cohort Analysis Overview

In general, cohort analysis is a customer behavioral analytics technique that analyzes the long-term behaviors of a subset (or segment) of your customer base. 

Cohort analysis is critical for most ecommerce businesses because most of your short-term analyses do not reveal your profitability’s true nature. 

Let’s examine this point with a relatively extreme case. 

For example, if you run a subscription business like Freshly or one of our clients, Realeats. Your customer’s first order might only be $50, against a $100 cost of acquisition, which makes the activity of acquiring this customer unprofitable at first glance. 

However, as time elapses, you recover your customer acquisition cost gradually. Over a three or even six month period, as more of those $50 orders start to come in on a week-by-week basis, the customer returns a significant amount of profit to your business. 

multipul order.png

The problem of your analytics without cohort metrics is that metrics such as CAC or revenue do not take into account your customers’ orders six months later. 

Those short-term reporting metrics make it very difficult for you to measure your actual per-customer profit if a significant part of your revenue comes from customers’ activities post-purchase - that’s why you have to analyze your cohort metrics in addition to metrics like CAC.

Key cohort metrics such as customer lifetime value and customer retention rate (or churn rate as an opposite measure) can help you understand how much value you are truly obtaining from your customers compared to their acquisition cost - an analysis that has enormous implications on your marketing strategy. 

However, as a flaw and a feature, cohort analysis is designed to be a long-term, lagging analysis. 

This means that, depending on the cohort elapsing time you choose, you can only get metrics of your cohort 3 or 6 months after the customer’s initial acquisition, making it incredibly difficult to leverage those metrics for rapid decision making. 

The consequence of this is that most companies we see are at most using cohort metrics for reporting purposes both internally and for investors but lack concrete plans to take actions from those metrics, which we will address and help you fix in this article. 

In the remainder of this article, I will briefly go over some recommendations on how you can take actions from your cohort metrics, with examples attached for more actionability. 

The key to taking actions from your cohort metrics -> metric conversion

When setting cohort metrics as a primary metric goal, the biggest challenge is that you will not get any feedback from your actions until a couple of months later, making it clunky and difficult to take rapid actions. 

Luckily, cohort metrics are always closely associated with many short-term metrics that can serve as an early indicator of how it will perform a couple of months later when the cohort counting is finished. 

Let’s use Customer Lifetime Value, or Revenue Per Customer, as an example. 

Customer Lifetime Value (aka CLV) describes the average revenue your customers provide to your company throughout their engagements with your brand. 

Ideally, this metric should describe all revenues provided by customers from their first order until the heat death of the universe. However, in practice, we need to narrow our definition to something like -> from their first purchase until six months after (may vary depending on your business).

In the case of CLV, it can be broken down into two categories of short-term metrics: frequency metrics and order value metrics. 

clv.png

Frequency metrics describe how often you are getting returning customers through the door and buying again. 

While most of the frequency metrics are cohort metrics by themselves (such as orders per customer), there are also plenty of short-term metrics to choose from, one of the ones we suggest is revenue from returning customers. 

Order value metrics describe how much revenue your customers are generating per order, and the most typical metric we recommend for you to track is average order value. 

Setting your eyes upon improving your frequency metric and order value metric will, in turn, increase your CLV over time, helping you continuously take actions while having a long-term goal to aim for. 

Now let’s go over quickly a couple of other popular cohort metrics, along with their short-term metric buddies that we can use to describe and track them while they are taking shape. 

Average Order Per Customers: frequency metrics and order value metrics just like CLV described above. 

aov.png

Revenue After First Purchase (an excellent measurement of retention efforts): frequency metrics such as revenue from returning customers and order value from returning customers. 

Refund Rate: Overall refunds (computed as refunds are processed, instead of when orders are placed, because the latter is a cohort metric and hard to compute). 

Churn Rate: Number of churns in the current period (regardless of when the customer was acquired). 

churn rate.png

A general rule of thumb of finding corresponding short-term metrics for their cohort counterparts is to find ways to compute the information described by the cohort metric without introducing the concept of cohort. 

For example, for the cohort metrics churn rate, we can ignore which cohort the churns we are experiencing are coming from and just blindly aggregate them all together without worrying about the concept of cohort (then you will have the metric “churns”, an excellent short-term metric to optimize for). 

With that said, you should almost never choose a “percentage” metric as your short-term metric, as you are essentially pitting the activity of people from the past with those from the present, which are not that comparable. 

Please note that the short-term metrics introduced here should not serve as replacements for the cohort metrics. This is because those short-term metrics are merely indicators of the cohort metrics’ performance and are vulnerable to many mathematical tricks and mirages. Therefore, you should always set up a dashboard tracking those long-term cohort metrics as they are coming in. 

Conclusion and next steps

In the next article of this series, we will move onto the second challenge of cohort analysis -> the need for qualitative research supplements to truly understand our customers’ behaviors. 

In that article, we will show you a way to use cohort analysis to identify your best customers and provide you with some background on how you can leverage qualitative research techniques such as user interviews and surveys to build a full customer profile of your best customers. 

See you next time!