AdMob Match Rate Dropped: What Should You Check?

Direct answer

First confirm that the decline appears in comparable, completed periods and that ad requests did not change in a way that distorts the rate. Then split match rate into the smallest useful segments: country, format, placement or ad unit, app version, ad source, and time. A broad decline and a one-country decline point to different investigations, even when the account-level chart looks identical.

Match rate is an upstream serving signal, not a synonym for fill rate or show rate. A request can be matched and still not become a shown impression. Use the counts behind each rate, mark the exact start of the decline, and compare that boundary with account restrictions, consent changes, app or SDK releases, mediation updates, source eligibility, and traffic mix.

Start with the metric relationship

Match rate = matched requests / ad requests

Read the numerator and denominator with the percentage. If ad requests rise rapidly while matched requests stay flat, match rate falls even though the number of matched requests did not. If both matched requests and the rate fall, the revenue risk is more direct.

Match rate is not fill rate

Match rate describes the share of AdMob ad requests that received a matched ad. Fill rate can be defined differently across mediation platforms and reports, so verify the product's numerator and denominator before comparing it with AdMob match rate.

Match rate is not show rate

Show rate concerns what happened after an ad was available: whether a matched or loaded ad produced an impression. A placement or app-flow issue can reduce shown impressions while match rate remains stable.

A rate is not a volume

Always keep ad requests and matched requests next to the percentage. Rate movement without the supporting counts can exaggerate a mix change or hide the segment that contributes most of the loss.

Establish the real start and size of the decline

Compare equal-length windows with the same weekdays and reporting timezone. Exclude partial days and allow for normal reporting latency. Record ad requests, matched requests, match rate, impressions, weighted eCPM, and revenue for both periods. This makes it possible to distinguish a percentage change caused by a larger denominator from a real loss of monetizable request volume.

Mark the last normal interval and first abnormal interval as precisely as the data allows. Add account or policy notices, consent changes, app releases, SDK or adapter updates, mediation configuration, app-ads.txt changes, traffic campaigns, and unusual events to the timeline. Timing creates a candidate, not proof; validate it with affected and unaffected segments.

Segment until the broad average becomes specific

Start with country and format because both can carry large differences in demand and serving behavior. Continue through placement or ad unit, app version, ad source, and hour or day. Keep the same request definition and time window in every split. If one segment contains most of the lost matched requests, focus there instead of trying to explain the account-wide average.

Rank segments by lost matched requests or lost revenue contribution, not percentage decline alone. A new low-volume placement may show a severe percentage drop yet contribute little to the total. Conversely, a modest change in a high-volume country can explain most of the loss. Blended rates are useful alerts, but segmented counts locate the work.

Check restrictions, consent, and traffic context

Review the Policy center and any ad-serving-limit messages without assuming every decline is a policy event. Confirm whether the affected apps, countries, or ad units match the scope of the notice. For EEA, UK, and Switzerland traffic, place CMP or consent changes on the timeline and compare consent-eligible segments consistently. Do not infer consent status from country-level rate movement alone.

Inspect unexpected request growth, acquisition-source shifts, and suspicious traffic concentrations. A surge of requests from a new source or region can lower the blended match rate even if established segments remain stable. Keep traffic-quality claims evidence-based: report the observable request and segment change first, then investigate validation or serving evidence through the appropriate platform tools.

Use releases and mediation changes as testable boundaries

If the decline begins near an app, SDK, adapter, or mediation release, compare old and new app versions over the same countries, formats, placements, and sources. Look for changed ad-unit mappings, initialization order, request timing, consent handoff, adapter state, or rollout coverage. A version-specific split is stronger evidence than a release date that merely happens to be nearby.

When mediation is involved, confirm which report defines each rate. A network's fill metric, the mediation platform's fill metric, and AdMob match rate may cover different stages or inventories. Use source-level requests, responses or fills, impressions, and revenue together. Continue with the mediation-update guide when the change aligns with a rollout, or the source guide when one source is isolated.

Diagnose fill after a mediation update · Diagnose one ad source that stopped filling

How to interpret the combined evidence

A broad match-rate decline with stable requests and a simultaneous serving notice supports a serving investigation. A decline isolated to consent-sensitive traffic after a CMP change supports a consent-flow comparison. A version-specific decline after an integration update supports an implementation check. A one-source decline supports a source-level review. None of these conclusions should be generalized beyond the segment in which the evidence appears.

If match rate is stable but impressions and revenue are down, stop trying to repair match rate. The loss may be in traffic, ad opportunities, load-to-show behavior, placement exposure, or timing. If match rate is down but weighted eCPM on served impressions is stable, quantify whether fewer matched requests explain the revenue gap before attributing the outcome to pricing.

Return to the complete revenue-drop flow

Turn the finding into a controlled next check

Summarize the evidence in one sentence that includes time, segment, and leading metric. For example: ‘Matched requests fell in rewarded ads for version 3.8 in two countries immediately after the adapter rollout, while other versions stayed stable.’ Then choose a check that could falsify that statement, such as validating mappings and request logs for the affected version against a control.

Define the expected signal and observation window before changing anything. Preserve one unaffected segment when possible, change one variable, and record the outcome in the same metrics used for diagnosis. If the test fails to improve matched-request volume or exposes a different boundary, revert the experimental change and follow the next supported branch.