Revenue Per Install in Apple Search Ads: Judging Installs by What They Are Worth
Installs are the number most of us stare at first. They are easy to see, they move every day, and they feel like progress. But an install is only a cost you paid to get a person into your app. Whether that cost was a good deal depends on what those people do afterward. A campaign can show a healthy install count and still lose money, and another can look expensive per install and quietly pay for itself.
This post walks through how to compare what you spend per install with what each install is worth to you, why the timing of that comparison matters, and how to avoid the common traps.
Installs are a cost, not a result
In Apple Search Ads, you bid a maximum cost-per-tap and pay an actual cost-per-tap set by the auction. Taps turn into installs at some rate. Divide your spend by installs and you get cost per install, often called CPI. That number tells you what you paid to acquire one download.
What it does not tell you is whether the download was worth the money. A $3 CPI is a bargain for an app whose paying users spend $40 a year, and a disaster for an app that earns $1 from a typical user. The install count is just volume. The question that matters is the value per install.
Working out revenue per install
The simplest version is revenue per install: the revenue attributable to a group of installs, divided by how many installs that group produced. Put the two side by side and you get a direct comparison.
An illustrative example, with made-up numbers: suppose an ad group spent 300 units of currency over a month and produced 100 installs. Its CPI is 3. Over the following months, those same 100 installs generated 360 in subscription revenue. Revenue per install is 3.60, and ROAS is 360 divided by 300, or 1.2. The campaign returned a little more than it cost, before your own costs and Apple's cut are considered. Take the platform fee and your other expenses into account before calling anything profitable.
The point of the exercise is that the same CPI can be fine or terrible depending on the revenue number next to it. Always read the two together.
Lag is the trap
Apple's attribution flow ties an install back to the ad tap through the AdServices attribution token, and that resolves within about a day. Revenue does not arrive on that schedule. A person who installs today may subscribe next week, or after a trial ends in a month, or never.
This creates a common mistake. If you look at a brand new campaign after three days, the revenue column is mostly empty, not because the keywords are bad but because the installs have not had time to show what they will do. Judging a fresh ad group on its early revenue makes every new campaign look like a loser.
The fix is cohort thinking. Group installs by the week or month they arrived, and compare each cohort only once it has had enough time to mature. Pick a window that fits your app: a free game with one-time purchases may settle quickly, while a subscription with a seven-day trial may need a few weeks. Write the window down so you apply the same rule every time.
Connecting revenue to keywords without per-keyword revenue
Apple does not report revenue per keyword. What you can do is link installs to the ad group and campaign they came from, then pull revenue for those installs from a tool such as RevenueCat or your own purchase records. The chain is install, then attribution, then purchase. At the ad group level this is usually good enough to make decisions.
To keep this clean, structure your campaigns so each ad group holds a related theme. If one ad group mixes a brand term, a competitor term, and a generic category term, its revenue per install is an average of three different audiences, and you cannot tell which part is working. Separate them and each number starts to mean something.
Set a target payback before you look
Most of the frustration in this area comes from deciding what counts as good after seeing the numbers. Decide first. A workable approach looks like this:
- Estimate the revenue a typical paying user brings in over a fixed window, such as their first 90 days.
- Multiply by your conversion rate from install to paying user, which you can measure from your own cohorts.
- That product is your expected revenue per install. Keep a margin below it for your costs.
- Your target CPI is that expected revenue per install minus the margin. Anything above it needs a reason to exist.
This gives you a ceiling. An ad group whose CPI sits comfortably under the ceiling and whose revenue per install is holding up is a keep. One that is above the ceiling after its cohort has matured is a candidate for a lower bid, tighter keywords, or a pause.
Watch for distortions
A few things will make the comparison lie if you ignore them:
- Trials and free tiers can inflate early install value if you count trial starts as revenue. Count only money actually received, or label trial starts separately.
- Refunds and cancellations reduce revenue after the fact. Net them out before comparing cohorts.
- Very small sample sizes swing wildly. A single large purchase on ten installs will make an ad group look brilliant. Wait for enough installs before trusting a ratio.
- Changing the product page, pricing, or paywall mid-window mixes two different products into one cohort. Note the change date so you can split the data.
A weekly habit that keeps it honest
You do not need a dashboard full of charts. Once a week, pick the ad groups with enough mature installs, compare CPI to revenue per install against your ceiling, and write down one decision for each: keep, lower the bid, tighten keywords, or pause. Then leave them alone for the next cycle so the next read is clean.
AdsBuddy is built around this kind of routine: it reads your Apple Search Ads data alongside revenue, then turns the findings into a short, prioritized list of changes. You review each one and apply it yourself, so nothing changes without your approval.
Takeaway
An install count tells you how much you bought. Revenue per install tells you whether it was worth buying. Compare them against a target you set in advance, give each cohort time to mature before judging it, and keep ad groups focused enough that the numbers have a clear meaning. Do that consistently and the question shifts from whether installs look good to which spend actually earns its keep.