Short answer: App ad revenue can fall when requests decline, fewer requests return ads, fewer returned ads display, or impressions earn less. Diagnose the requests-to-impressions funnel before changing pricing: higher eCPM can coexist with lower total revenue.
Diagnose the funnel before changing pricing
Follow requests, matched requests, impressions and estimated earnings. Each stage answers a different question. A load-to-display problem needs a different response from reduced ad availability or a change in audience demand.
Google’s reporting glossary describes the metrics and their report scopes. Match the definitions to the report you exported before combining counts. An app-level request count and one ad source’s impressions are not automatically compatible denominators.
The first goal is to locate the largest change, not maximize one rate in isolation. Increasing requests cannot fix a rendering problem, and a higher displayed eCPM does not guarantee more revenue.
Use the right denominator for each metric
Match rate equals matched requests divided by requests. Show rate equals impressions divided by matched requests. eCPM equals estimated earnings divided by impressions, multiplied by 1,000. The first two can be displayed as percentages but must be decimals in multiplication.
For a consistent scope, estimated earnings equals requests × match rate × show rate × eCPM ÷ 1,000. This identity is useful for reconciliation, not permission to mix currencies, periods, account levels or incompatible source metrics.
Google defines show rate as the proportion of returned ads displayed to users. eCPM is revenue per thousand impressions, not per thousand requests or users. Keep units beside every number in a worksheet.
Worked example: eCPM rises while revenue falls
The invented periods below share the same scope. Period A produces 100,000 × 0.80 × 0.75 × $4 ÷ 1,000 = $240. Period B produces 100,000 × 0.60 × 0.50 × $5 ÷ 1,000 = $150.
eCPM increased 25%, but revenue fell 37.5% because impressions declined faster. Restoring Period B show rate to 75% would produce 45,000 impressions and $225 if all other inputs stayed fixed. That is sensitivity arithmetic, not a promise that a display change leaves demand unchanged.
| Metric | Period A | Period B |
|---|---|---|
| Requests | 100,000 | 100,000 |
| Matched requests | 80,000 | 60,000 |
| Match rate | 80% | 60% |
| Impressions | 60,000 | 30,000 |
| Show rate | 75% | 50% |
| eCPM | $4.00 | $5.00 |
| Estimated earnings | $240 | $150 |
When match rate falls, inspect ad availability
Segment by app, ad unit, country, format, operating system and source or source instance. Compare dates before and after configuration changes. Locate the affected segment before concluding that the whole app has a defect.
Account alerts, source configuration, integration changes, consent implementation and pricing-floor changes are investigation candidates, not established causes from a dashboard alone. Check actual release history and logs.
Do not raise request frequency to disguise a lower rate. More requests can enlarge the denominator without creating genuine display opportunities. A change in country or format mix can also move the aggregate even if individual segments behave consistently.
When show rate falls, inspect the load-to-display path
A returned ad is not yet an impression. Trace opportunities from creation through load success, attempted presentation and impression recording. An abandoned screen, discarded preload or failed display attempt are hypotheses to validate with instrumentation.
Some loaded ads legitimately never display because a user leaves before reaching a placement. A universal 100% show-rate target is not justified. Compare versions and placements to find unexpected changes without forcing intrusive presentations.
Useful events include opportunity created, load requested, load succeeded or failed, display attempted, impression recorded and opportunity abandoned. Keep the same identifiers through the sequence so failures can be traced rather than guessed.
When eCPM falls, compare like-for-like segments
If volume, match rate and show rate are stable, eCPM is the main arithmetic explanation for an earnings decline. But an aggregate decrease can result from more impressions in lower-earning segments without any individual segment deteriorating.
Compare the same format, country, operating system and source over aligned periods. Then examine the impression distribution. Do not treat a weighted overall rate as the unweighted average of segment percentages.
Observed revenue-derived eCPM is not the same as a manually configured floor. The glossary distinguishes observed eCPM and third-party data provenance. Raising a floor should be evaluated through earnings, match rate and impressions together, not assumed to raise realized revenue.
Reconcile exports before testing one hypothesis
Align date range, timezone, currency, app and dimensions. In mediation, one app opportunity and multiple source-level attempts are not interchangeable. Adding source rows can create a denominator that does not represent unique app requests.
Calculate aggregate rates from aggregate counts. Total impressions divided by total matched requests is the aggregate show rate; an unweighted mean of small and large ad units is not. Preserve source export timestamps because reports may update.
Test one specific hypothesis with a controlled rollout where feasible. Define expected signals and guardrails: estimated earnings, impressions per user, crashes, retention and user experience. A higher rate is not success if the app becomes intrusive or harms long-term usage.
Diagnostic checklist and limits
These metrics describe outcomes, not definitive root causes. Small samples, changing audience composition, reporting updates and third-party data can distort comparisons. Estimated earnings are also not finalized cash receipts. No source here provides a universal good match rate, show rate or eCPM.
- Align period, timezone, currency and report scope.
- Compare request counts before interpreting rates.
- Recalculate match and show rates from counts.
- Reconcile impressions and eCPM to earnings.
- Segment by unit, format, country, OS and source.
- Connect changes to releases and configuration.
- Validate display hypotheses with logs.
- Monitor revenue and experience together.
Official sources and limitations
Policy sources consulted November 7, 2026. Platform terms, regional eligibility, payment screens and account notices take precedence over this guide. Example amounts are hypothetical, not guaranteed payouts. This is educational guidance, not personalized financial or tax advice.
Run your own numbers
Use your own statement amounts and measured rates instead of treating a calculator default as an official payment quote.
More monetization guides