Some online businesses, like SaaS companies, spend a lot of time thinking about their churn rate. (This is typically the percentage of customers who unsubscribe from their service in a given month.) For subscription businesses, a low churn rate can be the difference between life and death.

Companies with non-subscription business models, however, typically don’t think about customer behavior in terms of churn. E-commerce players, for example, think about long-term customer relationships in terms of conversion rates and repeat purchase rates. There is rarely a clear distinction between a “current customer” and a “churned customer” in these analyses.

That is, until recently. Every month we’re seeing more and more e-commerce leaders adding “churn rate” into their RJMetrics dashboards.

How can an e-commerce business measure churn if customers never “unsubscribe?” There are a number of approaches that make sense, and typically they depend on how your company plans to act on the data.

One of the most popular methods is to set a cutoff date after which, if a customer has not made a purchase, they are considered to have “churned.” The choice of this date can be arbitrary, or it can be influenced by things like time between orders, repeat purchase probability, and cohort analysis of historical data.

Once you have defined populations of “active” and “churned” customers, you can pursue a number of new analyses and strategies. Here are some tactics we’ve seen:

• Segment things like referral sources and product categories by percentage of customers churned. This can tell you more about where your most loyal customers come from and what they tend to buy, which can inform marketing and merchandising decisions. This is probably most valuable as part of a cohort analysis.
• Identify populations of customers who are “about to churn” and send them special promotions or offers to encourage a purchase. While these groupings are definitionally as arbitrary as your churn threshold, instituting this practice on a regular basis will ensure that you are consistently reaching out to new populations of at-risk customers (since customers from previous batches who did not purchase will have churned and those who did purchase will no longer be at-risk).
• Monitor the health of your business by tracking churn rate over time. Changes in the percentage of customers who move from “active” to “churned” in a given month can be an directional indicator of changes in customer loyalty or behavior.

Giving lapsed customers a definitive “churn” event changes the slow, unpredictable fade-away of an e-commerce customer into a sharp step function that can be monitored, quantified, and acted against. It can simplify goals and clarify vision.

For more on churn rate. check out www.churn-rate.com. To give this a try on your own data, you can try RJMetrics free for 30 days.

There are a lot of different ways to look at your data in RJMetrics, and we know that interpretation and understanding are just as important as calculation and visualization. So, I’m writing a series of blog posts where I will do a deep dive into some of our analyses and visualizations.

The first in the series is the lifetime revenue cohort analysis.

What does lifetime revenue cohort analysis mean?

This chart shows the cumulative spending per user for a period of time after they are acquired. Cohorts of users are split up by their acquisition month.

For example, the orange line above shows the average for users who were acquired in November 2011. The first data point means that in their first month, users who were acquired in November spent an average of about $200. The second data point means that by the end of their second month, these users had spent an average of about $240. Their average spending in month two was approximately $40 (240 – 200).

The different lines represent different cohorts of users. The green represents the users that were acquired in December, and the blue is users that were acquired in October.

Why is this important?

This kind of cohort analysis can be useful for several different purposes, but the most immediate benefit is often better customer acquisition decisions.

Many companies limit their marketing spend to channels that yield profitability on a customer’s first purchase. These companies will pay to acquire customers through a given channel as long as that their average first purchase yields more gross margin than it costs to acquire them. The problem with this approach is that it often results in an underinvestment in growth. If your competitors are marketing based on a deeper understanding of buying behavior, they will outgrow you.

The lifetime revenue cohort analysis helps you to understand the consequences of expanding your customer acquisition spending, and it provides an easy way to convey this to the rest of your team. If future customers behave like existing customers, then acquiring customers for a higher CPA will result in a predictable payback period. Depending on the cash position of the business, you can define what payback period you are comfortable with, find the relevant spot on the chart, and spend accordingly.

Additionally, you can use this analysis to see if you are getting better at onboarding, engaging, and generating revenue from the users you acquire.  For example, this cohort analysis is a great way to see if a free shipping promotion for new users resulted in repeat buyers or one time purchasers that never come back.

How will this vary for different business models?

For most businesses, the lifetime revenue cohort analysis chart will show a large amount of spending in the initial period and then increase more slowly over time.

That initial spike is due to the fact that customers are more likely to make their first purchase soon after they are acquired than at any other time. In cases where the acquisition event itself is a purchase, 100% of customers make a purchase in their first period. In cases where registration can happen before purchases, this effect is less drastic. As an example, Groupon would likely have a much lower initial jump than Amazon, because many of the people who sign up for Groupon don’t make a purchase right away.

Unless there are a high number of refunds, this chart will slope up and to the right after the initial jump. The rate of growth tends to decrease over time because customers are usually most active when they first sign up. This causes the average to drop because the number of people in the cohort stays constant regardless of how many come back to buy more.

In subscription businesses, the slope will decay less aggressively than in businesses where people make one-off purchases. Occasionally, a subscription business will actually have a slope that increases over time. It is rare to see this, but it is a great signal for the business when it happens. This does not mean that there are zero churning customers, but rather that upgrades for customers that stay more than make up for the customers that leave.

How is this calculated?

There are two simple inputs to this calculation: how many members are in the cohort (which never changes), and how much revenue those members generated in the given period.

To determine the members in the cohort, we count the number of users who were acquired in the period in question. An acquisition can be a first purchase, account creation, newsletter sign up, or some other event.

The revenue calculation is a bit more complicated.  We want to sum revenue for orders that were placed by members of this cohort and took place within a fixed time period from their acquisition date (ie the first three months).

Finally, we divide the revenue by the number of members in the cohort for each time period in the chart and add this value cumulatively over time.

What are the variations of this chart?

There are many different kinds of useful cohort analyses.  The most common variation is filtering by user acquisition source. For example, you might want to look at this chart for customers who came from organic search, paid search, or an affiliate program.

This will help you understand if the customers from one acquisition source are more loyal or valuable than another. Thrillist’s subsidiary JackThreads used this analysis to understand that one of their most expensive acquisition sources was actually its most profitable. After learning this, Thrillist shifted their marketing budget to the more expensive acquisition source and accelerated their growth. Read our case study about Thrillist and Jackthreads.

Another way to look at the data is with an incremental, rather than cumulative, data perspective.  This shows the incremental amount that an average user spends in each month after they are acquired.  This is useful for forecasting the amount of repeat purchases you will get from existing users.

We can look at this with other things besides revenue as well.  Some examples include margin as well as non financial metrics like invites, votes, or messages.

Conclusion

Lifetime spending cohorts are a powerful way of looking at at your customers’ buying behavior.  Stay tuned for more information on how to use and interpret your metrics.