App install campaigns are the clearest example of the wrong-metric problem. Optimise for installs and you will get installs — from people who uninstall within a week.
Optimise for what happens after
Both Google and Meta let you optimise towards in-app events. Pick the one closest to money you can generate at reasonable volume:
- Ecommerce: first purchase, or add-to-cart if purchase volume is too thin to learn on.
- Subscription: trial start early, then move to paid conversion once volume allows.
- Marketplace: first completed transaction, not signup.
- Content: a meaningful engagement — session length, third session, profile completed.
You'll pay more per install. Cost per retained user falls, which is the trade you want.
Attribution needs a real tool
App attribution can't be done with a pixel. You need an SDK-based measurement partner wired up before spending, or you'll have no idea which campaign produced which user. Set this up first, not after the first month disappoints.
A ₹12 install that uninstalls on day two costs more than a ₹90 install that buys something.
India-specific realities
- Device mix skews mid-range Android. Test onboarding on a cheap phone on 4G.
- App size matters. Large downloads lose people on mobile data before the install completes.
- Language. Onboarding in English alone leaks users in most consumer categories.
What to watch in app install campaigns
Day-1, day-7 and day-30 retention by source. A campaign with cheap installs and terrible day-7 retention is buying you nothing, however good the dashboard looks.
What an MMP actually gives you
Mobile measurement partners get skipped because they look like an extra line item before you have any users. Then the first month ends, three channels each claim the same installs, and there is no way to settle it.
An SDK-based attribution tool does three things a pixel cannot. It deduplicates installs claimed by multiple networks under a consistent rule. It ties in-app events back to the campaign that produced the user, weeks after the install. And it gives you cohort retention by source, which is the only view that tells you whether a cheap channel is actually cheap.
Install it before the first rupee of spend. Retrofitting attribution loses the entire learning period, and on app campaigns the learning period is most of what you paid for.
Reading cohorts instead of installs
Judge a channel on what a hundred of its users are still doing in a month, not on what they cost on day zero.
| Source | Cost per install | Day-7 retention | Effective cost per retained user |
|---|---|---|---|
| Channel A | ₹14 | 6% | ₹233 |
| Channel B | ₹68 | 31% | ₹219 |
| Channel C | ₹95 | 44% | ₹216 |
On the dashboard that most teams look at, Channel A is winning by a factor of six. On the number that pays your bills, all three are roughly equal — and Channel C is the one worth scaling, because its users are still there to sell to.
India-specific realities worth planning for
- Storage pressure is real. On mid-range Android, apps get uninstalled to make room. Keep the download size down; it affects retention more than most onboarding tweaks.
- Onboarding has to survive a bad connection. If your first screen needs a fast round-trip, you lose users in exactly the markets you are trying to reach.
- Phone-number sign-in beats email. OTP is familiar; email verification adds a step many users will not complete.
- Regional language matters more after install than before. English ads work widely; an English-only app loses people at the second session.
The wider version of this, including where Google accounts usually leak budget, is on our Google Ads page.
Questions people ask
Should app install campaigns optimise for installs?
Only at the very start, and only until you have enough post-install events to optimise on. Optimising for installs buys the cheapest installs available, which are usually from people who open the app once. Move to a real in-app event as soon as volume allows.
Which in-app event should I optimise for?
The earliest event that correlates with revenue and happens often enough to train delivery — registration, first search, or adding to cart. Purchase is the ideal target but is often too rare to optimise on in a small account.
Do I need an MMP?
If you are spending meaningfully across more than one network, yes. A mobile measurement partner deduplicates installs between networks and attributes post-install events to a source. Without one you are comparing self-reported numbers from platforms that each claim the same user.
What is a realistic cost per install in India?
It varies so widely by category and target audience that a single figure would mislead you. Work from your own numbers instead: what an engaged user is worth, what share of installs become engaged, and what that makes an install worth paying for.