Performance pricing means you pay when something happens. The hard part is choosing what that something is. Pick an event too early in the funnel and you pay for users who never matter. Pick one too late and you get so few conversions that nobody can optimise.
CPI: cost per install
You pay when a user installs and opens your app. It gives the most data, fastest, and suits apps with strong onboarding where most installs go on to register. It is also the easiest event to fake, so it needs the strongest fraud checks and a close eye on retention.
Use CPI when you need reach in a new market, your install-to-registration rate is reliable, and you can watch quality by source every week.
CPA: cost per action
You pay for an event after the install: a registration, a trial start, a first purchase or a verified account. Payouts are higher, but every conversion is closer to revenue and much harder to fake.
The right action is usually the earliest event that predicts revenue for your app. For a subscription app that is often the trial start. For a marketplace it is often the first purchase. For a finance app it is a completed verification.
Use CPA when you know your event-to-revenue rate, you want partners to share the quality risk, and you can send the event by postback reliably.
CPL: cost per lead
You pay for a submitted lead that passes your validation. It suits services and insurance, where a sales team turns enquiries into revenue. The model only works if you define a valid lead precisely, before launch.
- Which fields are required, and how they are checked
- Which locations and service areas count
- How duplicates are handled, and over what window
- How consent is captured and stored
Setting the payout
Work backwards from what a customer is worth. If one in five trial starts becomes a paying user worth a given amount, your maximum trial payout is a fifth of that, minus the margin you need. Start below the maximum, prove quality on a capped test, then raise payouts on the sources that earn it.