A live-streaming app running Google Ads app-install campaigns. The campaign was set up to a ₹20 target cost per install. I halved the target, then delivered at half of that again.
The campaign was configured to a ₹20 target cost per install. A target CPA is a ceiling — automated bidding will spend up to whatever number it is given, so a campaign set to ₹20 will tend to settle near ₹20. Inheriting that setting is not a strategy. The question was how far below it this campaign could actually run.
| Month | Installs | Spend | CPI | vs ₹20 target |
|---|---|---|---|---|
| January 2026 | 14,646 | ₹72,061 | ₹4.92 | −75% |
| February 2026 | 15,370 | ₹75,678 | ₹4.92 | −75% |
| March 2026 | 12,426 | ₹59,969 | ₹4.83 | −76% |
A target CPA is a ceiling, not a forecast. Automated bidding will spend up to whatever number you give it. Leave the opening target untested and that is roughly what you will pay.
App campaigns are signal-driven before they are creative-driven. Verify the install conversion feed first — everything downstream inherits its quality.
Inherited settings are assumptions, not constraints. The ₹20 target was in the account when I arrived. Testing it downward cost nothing and returned most of the result.
Efficiency has to hold to count. ₹4.92, ₹4.92, ₹4.83 across three consecutive months is a level the account can run at — not a good week.
Paying more per install than you need to?
Let's TalkFigures from the client's Google Ads ad group report, January – March 2026. Cost per install and install volume as reported by Google Ads. The ₹20 figure is the campaign's original target CPA as configured in the account, not an independently verified category benchmark. All amounts INR.