Cost per install vs cost per trial: which event should you optimize for?

Optimizing for installs gets you installs, not a business. Why the optimization event is the most important choice in your app campaigns, and how to make it based on signal volume.

Optimize your app campaigns for installs and you get installs. Optimize for trials and you get a business. The optimization event is the single most important choice in your setup, because the algorithm literally goes hunting for more people who fire that one event. The rule of thumb: pick the deepest event in your funnel that still has enough signal volume to learn on. For most apps that is the trial start, not the install.

Why is a cheap install a vanity metric?

Because an install has not earned you anything yet. Between the download and a paying user sit an onboarding, an activation moment and a paywall, and people drop off at every step. The campaign with the lowest cost per install often wins precisely because it finds the most fleeting users: people who tap on everything, collect apps and never start a subscription. The dashboard celebrates while revenue stands still.

The nasty part is that install optimization confirms itself. The algorithm finds cheap installers, your cost per install drops, and the campaign appears to get better and better. Only when you calculate through the funnel does the truth show: cost per trial and per paying user climb because the quality of the inflow falls. Look only at install cost and you systematically send your budget to the wrong users.

How do you choose the right optimization event?

The tradeoff is always the same: deeper in the funnel means more valuable signal, but less volume. The algorithm needs a healthy stream of events per week to recognize patterns; if volume drops too far, your campaign gets stuck in the learning phase and delivery turns erratic. So you do not simply pick the deepest event, you pick the deepest event your budget can feed.

• Install: the most volume, the least valuable signal. Only sensible for a brand new app or a small budget.
• Trial start or registration: the middle ground that fits most apps. Close enough to revenue, enough volume to learn on.
• Paid subscription or purchase: the most valuable signal, but too few weekly events for most budgets.

In practice, a healthy account grows through this ladder. You start broad on installs to build volume and data, move the event to trials once there is structurally enough volume there, and only test toward paid events when your spend allows it. It is not a one-time choice but a setting that grows with your budget.

The algorithm gives you exactly what you ask for, so ask for users, not downloads.

What does your optimization event mean for your creative?

More than most founders think. If you optimize for trials, your creative has to sell the trial, not the download. That changes the whole build: you show what the app delivers after the first week, you mention the trial period explicitly and you promise exactly what sits behind the paywall. An ad that suggests a free app while the paywall appears after the first session buys disappointed users and a low trial rate.

Congruence between ad and paywall is the cheapest optimization there is. Every promise in your creative gets checked further down the funnel by the user themselves. If the picture matches, the trial converts; if it does not, you pay for installs that vanish within a day. Treat your best performing paywall message as briefing material for your ads.

What do you do when trial signal is too thin?

The wrong reflex is falling back to installs and leaving it there. There are better options. You can concentrate budget in fewer campaigns so the signal per campaign increases, because ten fragmented ad sets all learn nothing. You can pick an intermediate event that fires more often but still predicts the trial, such as completing onboarding. And you can work on the funnel itself: every onboarding improvement raises not just your conversion but also your signal volume.

Meanwhile, measure what each campaign truly delivers at cohort level. Per weekly batch of users, track how many start a trial and pay, and put that next to your spend. That table matters more than any single platform number, because it shows which inflow still holds value a month later.

Conclusion

Cost per install and cost per trial are not competing KPIs; they belong to different maturity stages of your app campaigns. Pick the deepest event your signal volume can feed, build creative that makes the same promise as your paywall, and judge campaigns on what a paying user costs. A rising cost per install is entirely uninteresting as long as your cost per paying user falls.

We build these app campaigns daily: the event choice, the testing structure and the creative that attracts the right users instead of the cheapest ones. Not sure whether your setup is learning on the right event? Book a call and we will gladly take a look with you.