AdMob revenue drop
Why Did My AdMob Revenue Drop?
An AdMob revenue drop is the final symptom, not the diagnosis. Find the metric that moved first, isolate the smallest affected segment, and then choose one check that can confirm or reject the leading explanation.
Direct answer
Start by comparing two equivalent periods and decomposing revenue into impressions and weighted eCPM. If impressions moved first, investigate traffic and ad opportunities. If impressions stayed stable, split weighted eCPM by country, format, placement, ad source, and time. Match rate and fill rate help explain how requests became monetizable impressions, but they are not interchangeable and should not be treated as the same metric.
Do not begin with a floor change, mediation rebuild, or SDK rollback just because revenue is lower. Those actions can alter several signals at once and erase the evidence you need. A useful diagnosis identifies when the decline began, which metric led it, where it is concentrated, and what changed near that boundary. The flow below turns a broad revenue question into a limited next check.
Start with the metric relationship
Revenue ≈ Impressions × weighted eCPM ÷ 1,000
This is a diagnostic relationship, not a revenue forecast or guarantee. Use weighted eCPM for the same impressions represented in the revenue total; a simple average of country, format, or ad-unit eCPMs can give the wrong answer.
What changed first?
Use the first confirmed movement to choose a branch. A later change may be an effect rather than the cause.
| First signal | Likely direction | Follow this branch |
|---|---|---|
| Impressions fell while eCPM stayed near normal | Traffic, ad opportunities, show rate, or timing | Trace the impression loss before changing price settings. Diagnose falling impressions |
| Impressions stayed stable while weighted eCPM fell | Country, format, placement, source, period, or demand mix | Separate the pricing and mix components. Diagnose stable impressions and lower eCPM |
| Match rate fell before revenue | Serving restrictions, consent, request quality, segment mix, or integration | Compare matched requests with ad requests. Diagnose a match-rate drop |
| Blended eCPM fell while segment eCPMs looked stable | Country share changed | Rebuild the weighted country view. Check country mix |
| Fill weakened after a mediation update | Rollout, adapter, mapping, source eligibility, or configuration | Anchor the comparison to the release boundary. Check the mediation update |
| Revenue fell after floor changes | Higher price per served impression may not offset lost volume | Compare total revenue and fill, not eCPM alone. Review the floor change |
| One source lost fill or contribution | Source-specific eligibility, mapping, credentials, adapter, or demand | Keep the investigation at source and segment level. Diagnose the ad source |
AdMob Revenue Drop Diagnostic Flow
- Confirm the reporting window. Compare equal-length, completed periods with the same weekdays, timezone, apps, ad units, and currency.
- Ask whether impressions changed first. If they fell, trace traffic, ad opportunities, show behavior, placement exposure, and timing.
- If impressions stayed stable, compare weighted eCPM and then split it by country, format, placement, source, and period.
- Check match rate and fill rate with their underlying requests and matched or filled counts; do not use the terms interchangeably.
- Locate the smallest segment that still contains most of the loss, and rank segments by contribution rather than percentage change alone.
- Mark what changed near the start: release, CMP, serving notice, mediation, source, floor, campaign, event, or traffic mix.
- Choose one next check that can confirm or reject the leading explanation while preserving a control segment.
Read the outcome as a chain of signals
Revenue combines quantity and value. Impressions describe served ad volume, while weighted eCPM describes the revenue value of those impressions. Upstream, requests, match rate, fill rate, and show behavior explain whether available ad opportunities became impressions. Traffic volume, session depth, placement exposure, consent, serving status, and mediation eligibility can all affect that chain. Looking only at the final revenue line hides which link actually weakened.
Write a short timeline before opening dozens of reports: the last normal date, the first clearly abnormal date, app releases, consent or CMP changes, mediation changes, floor experiments, traffic campaigns, live events, holidays, and account notices. The timeline does not prove causation. It limits the candidates and tells you which before-and-after periods can be compared without mixing unrelated states.
Compare periods that answer the same question
Use equal-length windows with the same weekdays whenever possible. A Monday-to-Wednesday window should not be compared with a weekend, and a partial current day should not be compared with a completed day. Check reporting timezone, data latency, currency, and whether both periods include the same apps and ad units. If a holiday, sports event, school schedule, or user-acquisition campaign changed usage patterns, add a second reference period instead of forcing one misleading comparison.
Record both absolute values and rates. A match rate can decline while matched requests stay flat if requests grew quickly. eCPM can rise while revenue falls if impression volume collapses. Blended eCPM can fall even though every major country is stable when more impressions come from lower-value markets. The absolute numerator, denominator, and mix are what make a rate interpretable.
Find the smallest segment that still contains the drop
Start broad, then split in a fixed order: app, country, format, placement or ad unit, ad source, app version, and time. Stop when the decline becomes concentrated enough to test. If one rewarded placement in one country explains most of the loss, a global mediation rewrite is disproportionate. If every app, format, country, and source moves at the same boundary, look for account-wide serving, consent, reporting, or broad demand factors.
Use contribution as well as percentage change. A tiny placement can show a dramatic percentage decline without explaining the revenue total. Rank segments by the amount of lost revenue or impressions they contribute, then inspect the largest contributors first. This keeps the diagnosis tied to the business outcome without treating sample data or generic thresholds as a benchmark.
Read each branch without jumping to a fix
An impressions-first decline points toward fewer users, shorter sessions, fewer eligible opportunities, lower show rate, changed placement exposure, or timing. A weighted-eCPM-first decline with stable impressions points toward demand or a shift in country, format, placement, source, or period mix. A match-rate decline points upstream of impressions, but still requires segment evidence before you label it a policy, consent, or mediation problem.
A change that began immediately after a release deserves a controlled comparison, not an automatic rollback. Compare affected and unaffected app versions, placements, or sources. If only one source changed, inspect its mapping, eligibility, adapter state, credentials, and contribution. If a floor experiment preceded the loss, compare the revenue gained from higher prices with the revenue lost from fewer matched or served impressions.
Turn evidence into one next action
State the working diagnosis in one sentence: what changed, where, when, and what evidence supports it. Then choose the smallest next check that could disprove it. For example: ‘Revenue fell because rewarded impressions declined in one country after version 4.2; compare request and show behavior for versions 4.1 and 4.2.’ This is more useful than ‘AdMob performance is down’ because it defines an observable decision boundary.
Preserve a control segment and change one variable at a time. Document the expected signal, the observation window, and the rollback condition before acting. If the next check does not support the hypothesis, return to the table and take the next evidence-backed branch. A diagnosis is complete enough to act when the proposed change matches the affected scope and you can tell whether it improved the leading metric.
What not to change first
Keep the evidence stable until you know which branch explains the loss. Avoid broad changes that alter several metrics together.
- Do not raise or remove price floors across every country and format based only on blended eCPM.
- Do not rebuild mediation or disable multiple ad sources before identifying the affected source and segment.
- Do not roll back an SDK or app release without comparing affected and unaffected versions.
- Do not change placement frequency, consent flow, and serving configuration in the same test.
- Do not treat the illustrative sample as a benchmark, forecast, or customer result.
How to read it
Common interpretation
If impressions and upstream serving signals stayed stable while weighted eCPM declined across comparable segments, demand or auction value is a credible direction. If impressions, match rate, fill rate, country share, or one source moved first, follow that upstream change instead of calling the whole decline an eCPM problem.