Run install-optimised campaigns to seed a new app or a new channel with volume and early signal. Switch to event-optimised (in-app action) campaigns once you can feed the algorithm enough of the downstream event that actually matters — a registration, a subscription, a purchase. Installs are the starting line, not the finish. The real decision isn't "installs or events" — it's which event, and do you have enough of it for the algorithm to learn.
I run paid user acquisition across Meta, Google UAC and TikTok for a GCC fintech app, and this is the question that quietly decides whether a budget compounds or leaks. Here's how I make the call.
What's the actual difference between install and event-optimised campaigns?
An install-optimised campaign tells the ad platform "get me the cheapest installs." An event-optimised campaign tells it "get me the cheapest users who complete a specific in-app action" — sign-up, first deposit, subscription. Same ad account, different optimisation target, and the algorithm chases whatever you point it at.
That last part is the whole game. The platform optimises literally for the goal you set. Ask for installs and it finds people who install and vanish. Ask for registrations and it starts hunting for people who actually sign up — a smaller, higher-intent pool.
| Install-optimised | Event-optimised (in-app action) | |
|---|---|---|
| Optimises for | App installs (CPI) | A downstream event: reg, subscribe, purchase |
| Right KPI | Cost per install | Cost per in-app event |
| Best for | New apps/channels, seeding signal, volume | Quality, efficiency, scaling on the real KPI |
| Needs | Low event volume to start | Enough event volume for the algorithm to learn |
| Risk | Cheap installs that never convert | Starves for signal if the event is too rare |
Why are cheap installs lying to you?
A low cost-per-install feels like winning and often isn't. Install-to-registration rates can be brutal — on TikTok Android I've seen install-to-reg convert at just 5–10%. A campaign showing a great CPI while nine of ten "users" never sign up is not a good campaign; it's an expensive one wearing a disguise.
This is the trap the dashboard sets for you. CPI is an intermediate metric — cheap, fast, and only loosely connected to revenue. The number that pays your bills is cost per the event that signals a real user. On the fintech app I run, that's a completed registration; for a subscription app it's a trial or a paid start. Optimise to the install and you're optimising to the disguise.
The reframe that fixes most accounts: stop buying installs, start buying the event. Cheap installs are easy. Users who stay are the point.
When should you run install-optimised campaigns?
Run install-optimised campaigns when you don't yet have enough of the real event for the algorithm to learn from. Three clear cases: a brand-new app with no conversion history, a new channel you're opening from zero, or a market where your downstream event volume is simply too thin to optimise on directly. Installs are the widest, fastest signal available, so you use them to prime the pump.
There's a hard technical reason this matters on iOS. Apple's privacy framework — now AdAttributionKit (AAK), the successor to SKAdNetwork — only returns attribution data above a minimum volume. Campaigns running below roughly 20 installs a day frequently hit null conversion values: the data is suppressed entirely for privacy. The practical floor is around 100–150 installs per campaign per day to get meaningful post-install signal back. With ATT opt-in rates as low as ~14% globally, you no longer get clean user-level data for free — you have to earn enough volume to see anything at all.
So early on, install optimisation isn't a compromise. It's how you generate the volume that makes event optimisation possible later. Seed first, then sharpen.
When should you switch to event-optimised campaigns?
Switch the moment you can reliably feed the algorithm enough of the downstream event — as a rule of thumb, once a campaign can produce roughly 50 of the target event per week, it has enough to learn cleanly. That's when optimising to registrations or subscriptions starts beating optimising to installs, usually by a wide margin.
The gap is not subtle. On the same Android audience, my registration-optimised (Sales-objective) campaigns run $11–18 per registration, while pure install-led App Promotion campaigns on that same audience run $18–25. The best registration creative dips into single digits. Same people, same budget — the objective alone moves cost-per-real-user by 40% or more.
TikTok tells the same story even louder. When I moved TikTok off install optimisation onto In-App Action campaigns, install-led delivery had been running roughly 2x worse on cost-per-registration than the event-optimised version — bad enough that I killed the install campaign outright. The best iOS In-App Action subscription cost landed around $23.63, a number I'd never have reached buying installs.
A useful middle gear: Google UAC's "switch to in-app" campaign type starts install-optimised to gather signal, then hands the algorithm the in-app event to optimise toward. It's the transition built into the product — volume first, then quality — and on Android it's consistently my cheapest source of real registrations, at roughly $7 each.
What's the mistake that wrecks these decisions?
The single most common error is judging an event-optimised campaign by its cost-per-install. If you switch a campaign to optimise for registrations but keep reading the CPI column, you'll "see" a rising cost and kill a campaign that's actually working — because you're looking at the wrong number.
I've made bad kill calls this exact way: an In-App Action campaign looked expensive on install cost and got flagged, when its cost per registration — the metric that matched its objective — was healthy. The rule I now run on every review: before any keep-or-kill decision, confirm the cost-per-result column matches the campaign's objective. Install campaign → cost per install. Event campaign → cost per event. If the right column isn't in front of you, you don't have the data to decide yet.
One more thing the objective changes: what creative works. On my account, personality and influencer content performs near-target in intent-aligned registration campaigns but fails in install-optimised ones — the same asset, deployed against the wrong objective, dies. Objective isn't just a measurement setting; it changes which ads win.
A simple decision framework
Use this order of operations instead of guessing:
- Name the one event that equals a real user (registration, trial, first purchase). That's your true KPI — not the install.
- Can a campaign produce ~50 of that event per week? If no → run install-optimised to build volume and signal. If yes → optimise to the event.
- On iOS, check you clear the volume floor (~100–150 installs/campaign/day) or your post-install data goes dark.
- Match the metric to the objective every single time you review. Wrong column = wrong decision.
- Re-test creative per objective. Don't assume an install-campaign winner survives in an event campaign, or vice versa.
Do this and "installs vs events" stops being a philosophical debate and becomes a sequencing decision: installs to start, events to scale.
FAQ
What is an event-optimised campaign?
An event-optimised (or "in-app action") campaign tells the ad platform to find users who complete a specific action inside your app — registering, subscribing, purchasing — rather than users who merely install. You're billed and optimised on cost per that event, which ties spend to real value instead of downloads.
Do install campaigns still matter in 2026?
Yes — as a starting tool. Install campaigns are the right move for brand-new apps, brand-new channels, and low-volume markets where you don't yet have enough downstream events for the algorithm to learn. They generate the signal that makes event optimisation possible. They're a phase, not a destination.
Is cost per install a bad metric?
It's not bad, it's incomplete. CPI is a fast, cheap intermediate signal, but a low CPI with a 5–10% install-to-registration rate is a warning, not a win. Judge campaigns on cost per the event that represents a real user, and treat CPI as a diagnostic on the way there.
How many installs does iOS attribution need to work?
Under Apple's AdAttributionKit (formerly SKAdNetwork), campaigns below ~20 installs/day often return null conversion values, and ~100–150 installs/campaign/day is a safer floor for meaningful post-install data. Below that, privacy thresholds suppress the signal you'd optimise on. (Verify current thresholds — Apple adjusts them.)
What is Google UAC's "switch to in-app" campaign?
It's a Google App Campaign type that begins install-optimised to gather early signal, then shifts to optimising for an in-app action. It's a built-in bridge from volume to quality, useful when you want the algorithm to start broad and then narrow onto real registrations.
Should I ever run both at once?
Sometimes. On a new channel you might run install-optimised to seed while an event-optimised campaign warms up, then shift budget as the event campaign proves it can hold cost. Just don't compare them on the same column — each is judged on its own KPI.
Trying to work out whether your app budget is buying installs or actual users? I run a free App Growth Review — a short, no-pitch look at your live campaigns, your objectives, and where the money's leaking.
About the author — Ahmed Khoja is an app user-acquisition consultant with 10+ years in performance marketing, running paid growth for GCC fintech, marketplace and consumer apps across Meta, Google UAC, TikTok and Apple Search Ads, with a focus on MMP-based measurement and attribution.