Mobile user acquisition produces more numbers per campaign than almost any other channel, and most of them are noise. A standard dashboard will hand you impressions, installs, install rate, cost per install, day one retention, session length and a dozen creative level breakdowns before it tells you whether the money came back. The applovin ads metrics that survive scrutiny run in the opposite order: revenue per cohort first, then the cost of acquiring that cohort, then the behavioural numbers that explain any movement. Everything above that line is diagnostic detail, useful when something breaks, not the headline you lead with.
There is a structural reason this matters more here than on search channels. In app advertising buys installs, and an install is not a purchase. The gap between the click and the revenue can be days or weeks, and it varies enormously by title, by geography and by monetisation model. A campaign that looks expensive on day zero can be the best performing line in the account by day thirty. A campaign with a cheap cost per install can be quietly filling your title with users who open it twice and never return. Any measurement framework that stops at the install is measuring the wrong end of the funnel, and in the accounts we manage that single mistake explains most of the budget that gets misallocated.
Start with the cohort, not the calendar month
The most useful change most advertisers can make is to stop reporting by calendar period and start reporting by install cohort. A calendar month mixes revenue from users acquired two years ago with spend from users acquired last Tuesday, and the result tells you almost nothing about whether current buying is working. A cohort view groups users by the week they installed, then tracks what that group spent at day seven, day thirty and day ninety against what it cost to acquire them.
The applovin ads metrics worth putting on the first page
A defensible report is short. When we rebuild reporting for a mobile client, the first page usually carries five things and nothing else. Everything further down exists to explain a change in one of those five, and nobody reads it unless something has moved.
- Return on ad spend at a fixed cohort age, most often day seven and day thirty, so the comparison is like for like across weeks.
- Effective cost per install, split by the placements and creatives that actually carry volume rather than averaged across the whole campaign.
- Retention at day one and day seven, used as an early warning signal for audience quality before revenue data matures.
- Cost per meaningful event, where the event is the one that correlates with revenue in your title rather than a generic tutorial completion.
- Share of spend going to creative variants that are still improving, which tells you whether the account has fresh inventory or is running on fumes.
Notice what is missing. Impressions and click through rate are not on that list. They matter when diagnosing a delivery problem, and they are worth watching when a new creative launches, but they do not belong in a summary that a finance team reads. A campaign can double its click through rate and lose money, and if that number sits at the top of the page somebody will eventually mistake it for progress.
Early signals when revenue data is still thin
The awkward period is the first week, when you have spend but not enough revenue to judge anything. This is where most overreaction happens. Campaigns get paused on day two because the cost per install looks high, before a single cohort has had time to monetise. The discipline we apply is to pick two or three early proxy events, validate that they correlate with paying behaviour in historical cohorts, and then judge new campaigns against those proxies rather than against raw revenue.
Attribution, incrementality and honest caveats
Any conversation about applovin ads metrics runs into attribution sooner or later. Mobile measurement has become probabilistic in large parts of the market, and the platform reported numbers and your measurement partner numbers will not agree. They are not supposed to. The mistake is picking whichever source flatters the campaign that week.
Our approach is to name one source as the source of truth for decisions, usually the measurement partner, and treat platform reporting as directional. Then, periodically, test whether the spend is actually incremental by holding out a geography or pausing a segment and watching what happens to organic volume. Judged this way, applovin ads metrics stop being a dashboard artefact and start answering the question that comes up in every budget review: what would have happened if we had not spent this?
Building a report you can defend
A defensible report has three properties. It uses the same definitions every week, so trends mean something. It shows the cost side and the return side on the same page. And it says plainly what is uncertain, including the attribution caveats and the cohorts that are too young to judge. That last property is the one most agencies skip, and it is the one that builds trust.
When we hand over a mobile user acquisition report, the campaigns still inside their measurement window are marked as such. It makes the document less exciting and considerably more useful, and it means that when we do claim a result, nobody has to take it on faith.
Keep reading: Applovin Ads · Werbung