Turning retention into decisions: LTV, payback and budget

What does turning retention into decisions mean?

Turning retention into decisions means converting a cohort's retention curve into money terms, such as lifetime value, payback period and cost per retained user, so you can decide where to spend.

Lifetime value (LTV): the revenue you expect from one installed user over a set period, built from how many users are still active each day and what an active user is worth.

Payback period: the number of days until the revenue from a cohort covers what you paid to acquire it.

Which retention definition feeds which decision?

Classic retention feeds forecasts and LTV, cumulative retention feeds activation goals, and rolling retention should stay out of finance because its values keep rising as late returns arrive.

Rolling Day 7 for an older cohort has had more days to collect late returns than for a newer one, so it moves after you have booked it. That makes it unsafe for a budget or a forecast. The full comparison is in classic vs rolling vs cumulative.

Building a simple LTV from a retention curve

LTV over a horizon is the sum, across each day, of classic retention on that day multiplied by revenue per active user per day.

Hypothetical arithmetic, not real data. Say an active user is worth $1 per day, and a cohort's classic retention averages roughly:

• Day 0 revenue counts toward LTV even though Day 0 is not a return.
• Use exact-day classic values, or the sum overstates value: cumulative or rolling would count the same user on days they were not active.
• Past the last observed day you are extrapolating. Keep the horizon short, or state the assumption.

The spreadsheet and SQL formulas give you classic retention by day to drop into this sum. For revenue cohorts beyond the usual Day 0, 7 and 30 checkpoints, this guide to cohort analysis for performance marketers goes further.

Cost per retained user

A channel with lower Day 7 retention is not automatically worse. If it is cheaper, each retained user may cost less.

Cost per retained user: cost per install divided by the retention rate on the chosen day.

Hypothetical example:

Channel B retains fewer users but each retained user costs $2.50 less. Before moving budget, check that B's retained users are worth as much as A's, using revenue per active user by channel. The retention by channel guide covers how to read channels side by side.

Remember to count every cost of acquisition, including what you pay to measure it. This overview of mobile attribution pricing shows how measurement fees are usually charged.

Comparing payback by channel

Payback compares cumulative revenue per install with cost per install, day by day.

• Hypothetical: with the LTV curve above, a $3.75 install pays back around Day 21; a $7.50 install has not paid back by Day 30.
• Plot payback curves per channel on the same axes and look for which crosses cost first.
• A channel that pays back slowly but keeps rising may still win over a longer horizon. State the horizon you are using.

When to wait before acting

Young cohorts have fewer observed days, so decisions made early can flip.

• A cohort has a Day N value only once Day N has fully passed for it.
• Act on Day 1 for quick checks such as broken onboarding, but wait for Day 7 or later before moving meaningful budget.
• Check that the curve has started to flatten before projecting LTV. See reading flattening.

Guardrails

• Small cohorts: a few users swing the rate by many points. Pool cohorts or wait. See sample size.
• Mix shift: in the site's sample app, a made-up app, blended classic Day 1 falls about 7 points while organic and paid are each flat, because paid's share of installs rose. Always split by channel before acting.
• One-day spikes: a push or release lifts one calendar day across cohorts. Do not read it as a better cohort.