Field notes · 3 Mar 2026

How to read an LTV cohort without flattening the story

Ad-supported LTV is not IAP LTV with the labels sanded off. Day-0 ads, returning sessions, and paid UA mix do different work. When a chart offers one line, someone already chose which of those stories to hide. This is the reading method we use in Cohort LTV for Ad-Supported Apps.

Keep the first day visible

A large share of ad revenue on casual titles still arrives before the user has a reason to come back. If you average that into a D7 or D30 curve, UA will look cheaper than it is, or more expensive than it is, depending on who is holding the marker. We ask students to print D0 ad revenue as its own column and to refuse any dashboard that will not do the same.

Return is a second business

Returning sessions carry different fill, different formats, and often a different consent state. A cohort that “looks healthy” at D14 can be a D0 spike plus silence. That is not a moral failure. It is a product fact. Monetization changes that assume return will carry the week tend to punish the people who already paid attention on day one.

Paid UA is not a stain on the organic line

Blending paid and organic cohorts is how finance gets a smooth chart and how UA gets blamed for a creative test that never touched organic. In the workshop we keep at least two cohort families. If your sample is too small in Thailand alone, say so. A noisy split is still more honest than a quiet blend.

What “good enough” looks like

You do not need a perfect identity graph. You need a table you can explain in one sitting: who arrived, what they saw, what they paid in ads, whether they returned, and which of those levers you are willing to touch this month. Privacy-Safe Measurement for Ads is the companion lab when store language limits how far that table may go.

The Revenue Signal Lab uses this reading as homework before the brief. If your current LTV slide has one line, start there rather than with another tool.