How Many Installs Do You Need to Measure Retention?

"How many installs do I need to measure Retention?"

We get this question approximately 1000 times per year.

And frustratingly, our answer almost always needs to be customized to the stage of the project, the retention target, and the ideal tradeoff b/t cost and accuracy.

Here's a cheat sheet that can suffice in 90% of cases.

A simple lookup table.

Start by choosing your retention target.

Say, you are looking at D1, and hoping for 40%.

Let's suppose that

• Last time you measured, you got 35%,
• You want to prove retention increased to at least 38%
• You want to minimize UA cost

Reference the cheat sheet

• Step 1: Choose the 40% retention column (2nd one)
• Look for the relevant range (first row: 38%-42%)
• Step 2: Reference the same cell below
• Cell prescribes 2,305 installs. 2300 will suffice.

Here's a scenario with a different answer

Say you are focused on D1 and hoping for 35%

Let's suppose that

• You haven't done a retention test yet for this game
• You want to prove that D1 is at LEAST 30%
• You want to minimize UA cost

Reference the cheat sheet

• Step 1: Choose the 35% retention column
• Find the relevant range (3rd row: 29.8%-40.2%)
• Step 2: Reference the same cell below
• Cell prescribes ~320 installs

The cheat sheet will get you an approximation that should work 90% of the time.

If you need an EXACT answer, just ask your favorite LLM.

They are quite good at math.

Anyway, installs are expensive.

Acquire responsibly, my friends!