Attribution for subscription apps

Why subscription apps are harder to measure

The money arrives late. A user installs today, starts a free trial tomorrow, pays next week and renews next month. If your report stops at the install, a campaign that brings many trial starters who cancel looks the same as one that brings subscribers who stay for a year.

There is a second problem. Much of the revenue happens while the app is closed. A renewal is charged while the app may be closed, so the SDK inside the app does not report it. Subscription attribution depends on events that come from outside the app.

The five events to connect to an ad

Each row is an in-app event or a server event that should carry the same user ID, so the attribution tool can link it to the first open.

If you already use a subscription platform to manage purchases, connect it to the attribution tool before you write any event code. It usually sends the last three rows for you.

Which number to judge a campaign on

Use cost per paying subscriber, then revenue against spend. Cost per install and cost per trial are early signals. They are useful in the first days of a campaign and misleading after that, because trial-to-paid rates can differ a lot between channels and creatives.

Because renewals keep arriving for months, read revenue by cohort: group users by install week and watch how each group's revenue grows. That is how you see LTV build up and how you work out when a campaign pays back. A monthly plan and an annual plan pay back on very different schedules, so report them separately if you sell both.

What to send back to the ad platforms

Ad platforms are designed to optimise toward the event you report to them through a postback. Report installs and they look for people who install. Report trial starts or first payments and they look for people who do that.

Cancellations complicate this. Someone who starts a trial and cancels an hour later has still been reported as a conversion, and the platform learns from it. Airbridge Core includes Signal Hold, which holds a conversion postback on supported ad channels for up to 24 hours and drops it if a matching cancellation arrives.^[1] Whatever tool you use, check whether early cancellations and refunds can be kept out of what the platform learns from.

Set it up in this order

1. Give every user one stable ID and pass it to both the SDK and the subscription platform.
2. Connect the subscription platform to the attribution tool so trials, payments and renewals flow in.
3. Run one test purchase and confirm it appears under the right campaign with the right amount.
4. Switch each ad platform's optimisation event from install to trial start once volume allows.
5. Review cohorts monthly and compare ROAS at the same age for each campaign.

For iPhone campaigns, the guide to fine and coarse conversion values explains how a trial or a purchase is reported.