Why AdMob earnings dropped: a practical investigation

In Statistics → AdMob, keep the report mode, app, currency and filters fixed. Compare equal completed periods and note the account timezone. A partly imported day can make an otherwise healthy app look weaker.

Write down the first date of the change. Compare the same weekdays when possible, then check your release notes and AdMob change history for nearby changes. Timing is a clue, not evidence that a release caused the drop.

• Exclude the current unfinished day and check import coverage.
• Use the same app, countries and formats in both periods.
• Record the two date ranges and keep an unchanged baseline report.

Use earnings = impressions ÷ 1,000 × eCPM for the same report slice. If impressions fall while eCPM stays similar, start with traffic and ad delivery. If impressions hold steady but eCPM falls, compare the mix of countries, formats and demand sources.

Both can change at once. Use the calculator below to isolate one assumption at a time, then compare the combined scenario. Those scenarios explain the arithmetic; they do not predict the outcome of a product change.

For Network reporting, compare ad requests, matched requests and impressions. Match rate describes the share of requests that returned ads; show rate describes the share of returned ads that were shown. A lower show rate deserves a different investigation from a lower match rate.

For a match-rate decline, review demand availability and recent changes to floors or blocking controls. For a show-rate decline, ask the app team to inspect loading, expiry and display timing. Treat these as checks, not a claim that any setting is wrong. Do not encourage users to click ads as an optimization.

Sort the app, country, format and ad-unit tables by earnings. Compare each segment with its own earlier value; the largest current segment is not necessarily the largest contributor to the decline. Inspect one breakdown at a time so the same revenue is not counted repeatedly.

Choose a single reversible change with a clear hypothesis. Write down the expected metric movement, the comparison window and a user-experience guardrail before testing. Review earnings alongside impressions and eCPM; higher yield is not a win if the experience loses users.

• Keep an unchanged comparison group where practical.
• Compare enough complete days to avoid deciding on one unusually quiet day.
• Record results and remaining uncertainty before expanding the change.