What do we talk when we talk about Retention?

“We need to improve Retention!”. This is a usual claim I hear from General Managers and CMOs. When this happens, I usually take one step back: what do you really mean when you say you want to increase Retention?

Think about Retail, for example. At brick and mortar stores Retention doesn’t mean someone getting back to the store and walking around the halls. If you are a first-time shopper, any retail would like you to go back to the store and shop again. Is not only about bringing users back, but getting them back to do something.

The problem definition is important since most of the Retention metrics stop at the visit, without taking into account what happens during that visit. This is the pitfall of Retention curves and the usual DAU/MAU metric. Ricardo nails it in his response on this Quora thread:

DAU over MAU was born when people didn’t have access to better analytics. It’s somewhat of a nice metric to quickly eye-ball how people are engaging with your app or website but it is not conclusive of any kind of engaging behavior with it.”
Ricardo Vladimiro, Game Analytics and Data Science Lead @ Miniclip

So when we talk about Retention, we need to talk about users getting back to your app or site to perform your core action again (check this amazing presentation by Sarah Tavel on this point), whether this is listening to a song, watching a video or posting an ad for sale. The problem is that most of the Product and Marketing teams don’t have clarity on this —I have found that, most of the times, this problem comes out of having too many metrics on the scorecard.

Going back to the Retail example, Retention will mean to bring a user back to shop again, and maybe even increasing the average ticket size by sending tailored promotions (based on previous behaviors) or seasonal discounts in advance (in this sense, Urban Outfitter has been testing geo-based push notifications to influence physical transactions).

So, going back to the initial question, real Retention will be the one where a user gets back to your app or site to perform your core action. This is a cohort of your total user base and that’s why usual Retention curves are so tricky. Several analytics tools allow for segmentation of your user base, for example, the Retention of “users who favorited songs” vs “all users” in a music streaming app:

Demo data from Amplitude

Regardless the tools you use (third-party or in-house) for this analysis, here are some questions that will allow you to understand what we talk when we talk about Retention. I’ll use a hypothetical case of a video streaming company to make this clearer:

  • What is your core action? A video streaming company will value watching time as the main action, as this is the core of what they do.
  • What is the usual path to get to the core action? After account creation, the company found that not a lot of people perform a Search but rather discover its content by browsing the first screen.
  • What are the shared traits of people who churn? When analyzing their user flows, the team found out that, after watching three movies, users generally don’t watch any more content.
  • What are the main actions correlated with a higher Retention? They also found out that users with longer watching times usually create “Watch later” items after seeing content recommendations at the end of other shows.

It is important to mention that these correlations do not unveil any direct causal relationships, but will help you design smarter and more targeted experiments, in other words, you might want to explore something like this: “by increasing the number of times a user adds a title to the “Watch later” list, watching time for new users will increase by X%”. This kind of hypothesis will help you find the real triggers of long-term product retention and discover other possible avenues to take this further, like making changes on the “Watch later” feature or improving the suggested recommendations accuracy. The important takeaway here is that, with these insights, you will be able to understand exactly what Retention means for your company and, ultimately, turn them into actions for Growth.

UPDATE: check this article by mParticle: “using retention analysis, Netflix has identified how many episodes of a show you need to watch in order to become hooked. For instance, it takes just two episodes of Breaking Bad for 70% of viewers to get hooked, meaning that group of people will continue watching the remainder of the series to completion. Whereas it takes up to episode eight of How I Met Your Mother for you to become an avid watcher. Netflix then sends personalized notifications and alerts about when a user’s favorite show becomes available in the form of a new season or even other content they may enjoy.” Bam.