12 Measures of Retention: The What and The How
‘Retention’ is a term that commands attention in product management, yet it’s often surrounded by confusion. It’s a beloved metric due to its importance but frequently misunderstood as different professionals see it in various lights. When the conversation turns to retention, it’s rarely a unified diagnosis; instead, it’s a complex medical symposium where diverse perspectives converge, each providing a different diagnostic angle on what truly constitutes retention. I’ve consistently encountered at least four distinct viewpoints, offering a unique lens through which to understand this pivotal concept:
Retention in a broad sense is a synonym of customer habit to use the product.
Retention is a list of product improvements that you need to turn the users into habitues.
Retention is an indicator of whether you’ve achieved Product Market Fit or not.
Retention is the metric that helps to understand customer behavior.
Each perspective on retention is valid and insightful. My deep belief is that viewing retention as synonymous with habit formation is foundational. Yet, for the sake of this essay, let’s zero in on retention as a practical metric — the fourth viewpoint. In this sense, retention serves as a thermometer for your business’s health. It won’t offer diagnoses or pinpoint the causes of issues. Nor will it hand you treatments or solutions for better outcomes. What retention does provide is a clear indication of your product’s well-being: is it in robust health or suffering from a serious malady?
In any complex system, such as our body or customer behavior, it is impossible to assess the health status with a single number. Consequently, methods of calculating retention have become numerous and varied.
Frankly speaking, it’s not variety — it’s a mess. Thus, product managers add more words to definitions to get at least some clarity. To wrap your head around the level of chaos, let’s just list different metrics:
N-day, N-week, N-month Retention
Retention Curve
Rolling Retention or Unbounded Retention
Cohort Analysis
Week-over-Week, Month-over-Month Retention
Retention Rate
Active Users Retention Rate, New Users Retention Rate, Reactivated User Retention Rate
Action-specific, Feature-specific, or Subscription-specific Retention
And to make things worse, some concepts on the list can be combined with one another.
To discern which metrics are helpful for your product, let’s review them individually and examine their utility and limitations.
N-period Retention
The simplest and one of the most valuable metrics. Just measure the number of active users you have in the N-th period and divide it by the number at the beginning.
The trick here is to select the right period length, which depends on the natural usage frequency of your product. Reforge has a lot of great content on this topic. For example, for Facebook, the suitable period might be daily; for Netflix, weekly; and for Uber, it might be monthly.
The metric helps to understand the very beginning of the customer journey. Customers who passed the N-week mark and have been measured are no longer included in the analysis.
Curves
The N-week Retention metric doesn’t capture the entire customer lifetime; it’s merely a snapshot of a single timeframe. To create a holistic picture, we plot every N-th period on the graph, connect the dots, and receive one of the most common and valuable retention representations. If you were to ask a product manager at a bar, “Show me your retention,” chances are they would pull up a Retention Curve.
Of course, it can be charted based on the rolling (unbounded) version of the metric.
The most important parts of the Retention Curve chart are how high the retention is and whether it flattens. And if it does, it usually means you have achieved Product Market Fit.
The limitation of the Retention Curve is that it changes slowly, so it can’t provide frequent feedback. It’s hard to understand if your retention is improving just by looking at the curve.
Rolling or Unbounded Beast
A lot of products have long or unstable natural frequencies. For fashion ecommerce, travel & tourism, and the like, even second-month retention makes no sense as such categories aren’t consumed that often. Switching to quarters or years doesn’t help either because the metrics take too long to update.
So, these businesses improved N-month retention with the ‘rolling’ word. It basically means that the customer returned in the N-th month or any after. Consequently, the rolling retention rate is always higher than just counting the N-th period, since it includes all later returning customers too.
‘Rolling’ is the most common word for this method, but it doesn’t reflect the actual calculation. There is no rolling average or similar math here. Therefore, Amplitude, we at Reteno, and other experts advocate for an ‘unbounded’ term to name these metrics.
Cohorts
To add more dynamics to the retention curve, we can split our customers based on when they began using our product. So, we have a Cohort Analysis.
Retention Rate
Retention Rate is different from all metrics above. Most people better understand mathematical concepts through visual aids like charts, lines, and bars. However, if you’re one of the rare individuals who prefers algebra, this approach is tailored for you. The formula is indeed simple:
But the cases where you can use it are limited. Such calculations work only for continuous activity. For example, you have a subscription app and measure the Retention Rate by payments. In that case, it works well because the number of active subscriptions has a continuous nature and is known on any given day.
If you measure retention by target actions, the ‘at the end’ criteria isn’t applicable; you can only use ‘during.’ For instance, what do you mean by ‘Number of users at the end of February 2024?’ Does it refer to users who performed something at the last minute of February? Last hour? Last day? Last week?
This is a tricky question, to answer which we need one more concept — User Lifecycle. Let’s talk about it in the next section.
Pirate-like ‘RRs — AURR, NURR, RURR
When we add duration and time dimensions to the Retention Rate, we end up with the metrics that I personally love most. I am inspired by the elegance, simplicity, and meaningfulness of the Active User Retention Rate (AURR), New User Retention Rate (NURR), and Reactivated Users Retention Rate (RURR).
We must first segment the users to calculate them. For simplification, let’s focus on one of three — the Active Users. Every app will have its own definition. From a business perspective, we want Active users to represent customers who habitually use the app. So, we should measure the consistency of use over some periods. For example, at Zynga, Active Users are defined as those who come back this week if they have used the product each of the past two weeks. At Duolingo, Current Users are those who engaged today and at least one other time in the prior 6 days. I’m going to write a separate essay about selecting the correct period for AURR, so stay tuned 👀.
I recommend starting with the following user lifecycle segmentation with five segments. Here’s an example tailored to a fitness app with weekly natural use frequency:
Action-specific Retention
In all the metrics above, we used some kind of Active User definition. But what kind of activity makes them so? This question ties back to whether users are genuinely habituated to using our app. Simply measuring app opens isn’t sufficient. We need the customers to get value from the product: to play a game, to work out in a fitness app, to exercise for learning, etc.
Any previously mentioned metric should be calculated in an action-specific manner to gain deeper insights into customer behavior and improve business decisions. For instance, when calculating active users, ensure it is based on relevant events, such as a completed session in a meditation app.
Sometimes, ‘action-specific’ is referred to as ‘feature-specific.’ In this context, it means measuring retention for a bucket of users to understand the impact of deploying a new feature.
100% Retention
My doctor friends say, “There are no healthy people, only those who haven’t been examined.” Similarly, retention can’t reach 100%. However, its measurement can tell you how well you’re doing and how effective the treatment is. In a constant chase to create a habit, retention tracking gives you the speedometer to understand your pace. Just select the right metrics.