Retention after the install is a different species of number

Install is a store event. Return is a relationship the first session either earned or failed to earn. Those two facts are routinely mashed into a single tile labelled “retention,” which is how a product team ends up diagnosing character when they should be diagnosing the opening ten minutes.

In the Product Signal Atelier we treat the first thirty days as a short essay. The subject is the job the app claimed on the store listing. The verb is whatever the first session actually allowed the person to finish. The missing object is usually a reason to come back that was never spoken aloud — only implied by a paywall or a notification permission prompt.

Day one is not a personality

When day-one return sits at a number the board dislikes, the conversation often turns to “quality of users,” which is a polite way of blaming acquisition. Sometimes acquisition is the problem. More often the first session asked for an account, a notification, and a tour before it did the job. App Analytics can show you the drop. It cannot, by itself, decide that the tour was a courtesy. That decision is editorial.

Write the first session as a sequence of permissions the product requested. If the sequence is longer than the job, the curve will look like abandonment and feel, from inside the company, like mystery.

Denominators that keep you honest

A retention plot with an unspecified denominator is decoration. Say whether you mean return among people who completed a first meaningful action, or among everyone who created an account, or among store installs that never opened. Those three curves are not cousins. They are different animals wearing the same axis labels.

We ask students to annotate the chart in prose underneath the image, in the same document, so a future lead cannot screenshot the line without the sentence. It is a small clerical habit. It prevents a surprising amount of theatre.

What this essay will not do

It will not give you a benchmark. Benchmarks for consumer apps are a way of importing someone else’s first session into your argument. If you need a comparison, compare last month’s first session to this month’s, after you changed one request. That is slower than a slide titled “industry D1.” It is also closer to measurement.

The four-week intensive Retention After the Install is the compressed version of this reading. The twelve-week atelier spends a full module on it because most catalogues are not ready to support the sentences until week four.