The AdMob report shows a dozen metrics, but only a handful actually explain your revenue. If you can read those few and how they relate, you can diagnose almost any change. Here is the minimum set, what each one means, and the relationships that turn raw numbers into an explanation.
The metrics that matter
- Estimated earnings. Your revenue. Everything else exists to explain this number — never read it alone.
- Impressions. Ads actually shown. Half of revenue is just this count.
- eCPM. Earnings per 1,000 impressions — your effective price. The other half of revenue. Earnings ≈ impressions × eCPM ÷ 1000.
- Ad requests. How many times your app asked for an ad. The top of the funnel.
- Match rate. Matched requests ÷ ad requests — the share of requests that got an ad to potentially show. Low match rate = a fill problem.
- Show rate (impressions ÷ matched requests). Of the ads that were available, how many you actually displayed. Low here = an app/timing problem, not a demand problem.
- Match rate vs. impression eligibility. Watch the funnel: ad requests → matched → impressions. A drop at any stage points to a different cause.
The one relationship to internalize
Revenue is impressions × eCPM. So whenever earnings move, your first question is always: which factor moved — count or price? Put estimated earnings, impressions, and eCPM side by side for the period in question vs a stable baseline. If impressions moved, walk the funnel (requests → match rate → show rate). If eCPM moved, it's a pricing/demand story (floors, sources, geo mix). This single split tells you which half of the report to investigate and saves you from reading everything.
Use breakdowns, not just totals
The single biggest mistake is reading only the top-line. The total is a blended average that hides the cause. AdMob lets you break every metric down by country, ad unit, ad source (if mediating), format, platform, and app. A change almost always concentrates in one slice. Reading the breakdown — not the total — is the difference between "revenue is down" and "US rewarded eCPM on app X fell because a source dropped."
Compare the right windows
Don't eyeball the whole quarter. Compare a short recent window (yesterday, or the last 7 days) against a stable baseline (the prior 14–28 days). Differences hide in the diff, not the level. Also note your app release dates on the timeline — a metric that breaks exactly when a new build shipped is an app-side cause, not a market one.
What "normal" noise looks like
Daily eCPM wobbles a few percent; weekends differ from weekdays; end-of-month and seasonal demand shifts are real. Don't react to a single day. React to a sustained move that survives the count-vs-price split and concentrates in a breakdown.
Doing the count-vs-price split and the per-breakdown scan by hand, every time, is the chore Mediation One removes: upload your AdMob CSV and it does the decomposition and names the slice responsible. The free audit is one CSV upload — no SDK, no signup, nothing stored.