Incremental installs,
measured honestly.
Some people would have downloaded your app anyway. The only installs worth paying for are the extra ones. Here's how Ads Manager measures them, and where the method stops being trustworthy.
The idea in one paragraph
Look at how many installs you got per day in a campaign's countries during the week before it started. That's your baseline: what happens without the ad. Once the campaign is running, anything above that baseline is lift. Divide what you spent by the lift and you get the number that matters: the cost of each install the ad actually created.
extra installs = Σ (installs/day after launch − baseline/day)
lift = after/day ÷ baseline/day − 1
cost per extra install = spend ÷ extra installs
A worked example
You launch a campaign in Canada. In the 7 days before, Canada averaged 5 installs a day. Over the next 14 days it averages 12 a day, and you spend €70.
- Total installs: 12 × 14 = 168. Naive CPI: €70 ÷ 168 = €0.42.
- Extra installs: (12 − 5) × 14 = 98. Lift: +140%.
- Cost per extra install: €70 ÷ 98 = €0.71.
The naive number flatters the campaign by crediting it with 70 installs you'd have had anyway. Try your own numbers in the cost per install calculator.
When not to trust the number
A before-and-after comparison is not a controlled experiment. Ads Manager shows lift only when it can be read cleanly, and flags it when it can't:
- Overlapping campaigns. If another campaign was already spending in the same countries during the baseline week, the baseline is contaminated. The campaign is flagged so you don't trust a muddy number.
- Too little history. Lift needs at least three baseline days with data. If the campaign started before your data range, there's no baseline to compare against.
- Seasonality and features. A holiday, an App Store feature or a press mention during the campaign will look like lift. Check the timeline for spikes that started before the spend did.
- Tiny numbers. Going from 1 to 2 installs a day is +100%, and mostly noise. Give it volume and time.
Why country-level matters
Comparing worldwide installs before and after hides most of the signal. A campaign in Portugal won't show up against your US organic traffic. Ads Manager restricts both the baseline and the after period to the countries each campaign targets, using the storefront split in Apple's sales reports. That's what makes the method usable for small apps with modest budgets.
What it tells you to do
Campaigns with a clean, strong lift and a low cost per extra install are the ones to scale. Campaigns that spend a meaningful amount with installs stuck near baseline get called out as candidates to pause or refresh with new creatives. The Advisor turns those findings into a ranked plan.
Questions
What's the difference between CPI and cost per incremental install?
CPI divides spend by all installs in the period, including people who would have found you anyway. Cost per incremental install divides spend only by the installs above your baseline, so it's always higher and always more honest.
Why a 7-day baseline?
It covers every weekday once, so weekly patterns cancel out, while staying close enough to launch that trends haven't shifted much.
Is this as good as a geo holdout test?
No. A holdout with matched control regions is more rigorous. Before-and-after lift is the practical version for teams that can't afford to hold markets back, and it's surprisingly informative when campaigns don't overlap.
Know what's working.
Free for one app and one ad account. Every Pro feature free for 14 days.