Skip to content
All field notes
Measurement
Published
Nov 4, 2026
Reading time
11 min read

AdMob House Ads Not Serving? Audit Priority and Eligibility

Diagnose AdMob house campaigns that show paid ads or no impressions. Separate 100% priority, creative eligibility, request evidence and campaign reporting.

AP
AdProof Research
Independent measurement desk
A cross-promotion ad card reaches a priority gate while separate eligibility checks direct an unavailable card toward a paid-ad alternative.
Campaign priority, request eligibility and counted delivery are separate checkpoints. Conceptual illustration without app measurements.
In this guide (11 sections)

An AdMob campaign set to 100% of impressions receives first priority; it is not guaranteed to supply every impression. Investigate campaign eligibility for the actual request, then establish which ad loaded and whether it displayed. Raising a floor cannot substitute for that evidence. Google's campaign goal reference explicitly distinguishes priority from universal delivery.

An older cross-promotion question describes a single interstitial creative, a 100% impression goal and continued third-party ads. A separate mediated-house discussion describes house delivery stopping while paid ads continue. These are different observations, not proof of one platform defect. The second thread also mixes house and direct-sold terminology, making the configured campaign type essential evidence.

The useful deliverable is a request-specific finding: the intended campaign, applicable goal, eligibility checks, loaded response, display outcome and comparable report slice. Keep an unresolved selection reason unresolved rather than claiming a hidden auction decision from a percentage.

Identify the campaign contract before changing settings#

Write down the actual campaign type and goal from the saved configuration. House campaigns promote your own apps or websites; direct-sold campaigns represent an advertiser. The campaign setup guide separates these choices from the selected ad units and advanced targeting.

Do not use “house ad” as shorthand for any ad with no platform charge. A campaign promoting another owned app, a direct-sold advertiser and an app-owned fallback screen have different delivery evidence. Likewise, the app displaying the promotion is the placement; it is not necessarily the app being promoted.

Configured goalRelevant contractFirst diagnostic question
Percentage of impressionsFirst priority for the specified shareWas this request eligible for a matching creative?
Mediated adsCompetition in the mediation chain, using backfill optimization or manual eCPMWhat response path was selected?
Number of impressionsA dated impression goal with delivery pacingAre the selected period and eligible inventory consistent with the goal?

These distinctions come from Google's goal definitions. A manual eCPM used to position mediated house demand is not the same setting as a percentage goal or an ad-unit floor. Its monetary label is not a charge for mediated house ads. Save all relevant settings before proposing a change; otherwise the investigation loses its baseline.

For your campaign

Check eligibility against the actual request

A 100% priority setting does not ensure delivery on every impression. Discuss what request and serving evidence is available.

Discuss a measurement question

A short inquiry to scope relevant options; no campaign access is needed.

Separate campaign status from individual ad status#

Check the campaign's dates, status, selected units and available ads. Google's campaign-status reference distinguishes active eligibility from paused, future, ended and problem states. A campaign without ads or ad units requires a different response from one that is active but not selected on a particular request.

Then inspect each creative through the Ad dimension and Ad status attribute. The ad-status reference distinguishes Enabled, Approved (limited), Paused and Error. Enabled means available to serve. Approved (limited) has restrictions on where or when it can show. Neither label proves a compatible creative served the device under investigation.

Keep three separate records:

  1. What the campaign configuration permitted at the observation time.
  2. What the particular creative's status permitted.
  3. What actually happened for the tested request and display opportunity.

Do not confuse campaign Ready with the app-readiness state. If the unresolved gate concerns app verification or account approval, use the AdMob approval and readiness checklist. A badge at one level does not validate every other level.

Check eligibility against the actual placement request#

Google lists review failure, targeting mismatch, reached frequency caps, incompatible creative sizes and child-directed requests as reasons paid ads can appear despite a 100% goal. See the percentage-goal exceptions. The practical task is to connect one exception to evidence, not simply tick a generic checklist.

Create a private comparison for the released build:

EvidenceCompare withPreserve an unknown when
Installed app and actual ad-unit IDCampaign's selected placement unitsOnly the repository's intended value is known
Device platform and requested formatPromoted app and usable creativeNo matching request trace is available
Request contextEffective targets and applicable restrictionsThe relevant context cannot be observed lawfully
Ad status at the timeCreative attached to this campaignOnly today's status survives
Displayed responseIntended campaign and placementThe ad was classified only by appearance

The Campaign FAQs identify further setup checks: correct Android/iOS pairing, linked Google Ads billing, excluded promoted inventory and self-promotion within the same app. They also distinguish zero installs caused by absent conversion tracking from missing impressions. Campaigns cannot serve on apps targeting children or traffic from Google accounts registered to child users. Preserve those protections; do not change audience declarations to make an experiment pass.

Inspect targets as configured, including device language, geography, connectivity and operating-system restrictions where applicable. Google's setup documentation explains these controls. The device you used for an informal check may fall outside the campaign's intended audience even when another device qualifies.

Prove which response loaded before calling it third-party leakage#

A screenshot can establish that something appeared, but usually cannot identify the complete serving route. Retain the time, app version, platform, unit, format, load result and displayed outcome. Connect them with a short diagnostic correlation label, without collecting unnecessary personal identifiers.

For the documented Android SDK path, response information exposes the response identifier, loaded adapter and individual adapter responses. Record the loaded source and source instance where available. The adapter list describes participating responses; do not classify every listed adapter as a displayed ad. Also, a null adapter error alone does not prove success: the reference says it can mean an unattempted source.

Use the SDK family and platform actually installed. Do not copy Android method names into an iOS or Flutter investigation. Keep diagnostic records private and redact credentials or unnecessary identifiers before sharing support evidence.

Trace loading and display separately. If a response loaded but the user left before the intended transition, that is not proof the campaign was ineligible. If the app never requested the intended unit, campaign tuning cannot answer that missing application handoff. Record the first demonstrated boundary instead of assigning every missing impression to demand selection.

Backfill is not an automatic retry after every paid-ad failure#

For mediated house backfill specifically, Google's FAQ says a paid ad selected first cannot be replaced after the fact by a mediated house ad if it later fails. A slow waterfall can also miss the house opportunity. This is a scoped explanation of that path, not a universal rule for every percentage campaign or app-owned fallback implementation.

Separate these outcomes in the incident record:

  • No usable campaign response was selected.
  • A different response was selected and displayed.
  • A response was selected but failed later.
  • The app abandoned the opportunity before display.

They require different evidence and potentially different owners. Do not describe the last three as an identical “no fill” cause. Avoid repeatedly recreating campaigns, changing several targeting controls and raising floors together: even a later recovery would not identify which change mattered.

Compare Campaigns and Ads Activity without inventing a common denominator#

Save an explicit, completed date range, timezone, extraction time and filter set for every report. In Campaigns, Google's reporting reference provides Placement, Date, Ad, Country and Ad format. Placement identifies the hosting app. Its placement-name and store-ID filters require the Placement dimension.

Use the Ads Activity report for the corresponding app, unit, platform, format and source scope. The full-list source dimension separates sources within AdMob Network, which contains both Google and third-party demand. A high-level AdMob Network row therefore does not establish that only Google demand supplied the ads.

Prepare a comparison sheet containing the report name, exact metric, included sources and every retained dimension. Mark dimensions unavailable on one side; do not imply the Campaigns view has an ad-unit breakdown merely because another report does. Narrow the comparison to supported common scope and retain the unresolved granularity.

Source-level request rows also count calls as mediation progresses; adding them can produce multiple source requests for one ad-unit opportunity. Keep that denominator separate from campaign impressions. A share calculated from mixed request and impression populations cannot establish what fraction of eligible campaign opportunities served.

A hypothetical audit: two valid percentages, different questions#

Suppose a deliberately instrumented observation ledger contains 1,000 distinct selected opportunities in one app, format and completed period. The figures below are hypothetical and are not AdMob report results. Assume each selected opportunity receives exactly one mutually exclusive outcome classification:

Observed outcomeOpportunities
House creative displayed650
Paid creative displayed250
No ad displayed, with complete observation60
Observation incomplete40
Total selected opportunities1,000

House displays are 650 / 1,000 = 65% of selected opportunities. Among the 900 opportunities with an observed ad display, house represents 650 / 900 × 100 = 72.22%, rounded. Incomplete observations are 40 / 1,000 = 4%. None of these is automatically the platform's campaign-delivery percentage, match rate or show rate.

The ledger has not established how many of the 1,000 requests were eligible for the campaign. It also has not established whether the platform counts each display identically. Do not label the 350 remaining opportunities “lost house impressions” or turn them into a revenue claim.

Now inspect a small, deliberately chosen set of paid-display traces. If one definitively used a different unit than the campaign targets, that establishes a configuration mismatch for that trace. It does not explain all 250 paid displays. Record the demonstrated count and leave the rest unallocated until evidence supports another finding.

Keep estimated cost separate from cash and serving acceptance#

The Campaigns estimated-cost definition describes estimated revenue forgone by showing a house or direct-sold campaign ad instead of revenue-earning mediation or bidding demand. Google says AdMob campaigns are free. This field is not a platform campaign invoice, a verified counterfactual clearing price or proof of recoverable loss.

If the confusing comparison is callback revenue versus Firebase, use the separate ad-revenue collection audit. Keep delivery selection, revenue measurement and cash settlement as different acceptance records. A change that produces more cross-promotion displays does not by itself prove more installs, retained users or profit.

Close the investigation with a bounded acceptance checklist#

Use authorized test traffic to check the app's wiring. Google's Android test-ad guidance distinguishes demo units from test-device use of your own units, requires network-specific mediation testing and states that test-ad activity is absent from account reports. A demo response does not validate your campaign's production eligibility or reproduce its commercial selection mix.

When audio behavior is part of acceptance, verify AdMob volume and mute state on the actual format and source separately from campaign eligibility and account reporting.

Before closing, retain:

  • Campaign type, goal, schedule and the original configuration snapshot.
  • Placement identity, promoted app, creative status and relevant eligibility evidence.
  • A versioned request/load/display trace using the supported SDK diagnostics.
  • Comparable report exports with explicit unavailable dimensions and unknown outcomes.
  • A demonstrated cause where established, authorized change and subsequent retest.
  • Remaining unexplained paid displays or missing impressions, with the next evidence owner.

For campaigns already measured with AdProof, keep a supported campaign-event or aggregate export with matching dates, timezone and filters alongside independently obtained app evidence. The anonymized FMCG in-app count-once case concerns delivery reconciliation, not an AdMob campaign repair. Discuss the available measurement evidence to establish supported scope; AdProof does not expose private AdMob selection decisions or guarantee house delivery.

Frequently asked questions#

Does a 100% impression goal block every paid ad?

No. It gives first priority, while request and creative eligibility still matter. Save the actual goal and investigate a specific paid-display trace before assigning its cause.

Why can a Ready campaign show zero impressions?

Campaign-level readiness is only one checkpoint. Inspect individual ads, placement and promoted-app compatibility, applicable restrictions, request evidence and the selected reporting scope. Do not confuse zero installs with zero ad displays.

Should I raise the ad-unit floor until only house ads appear?

First establish the campaign goal and the failed boundary. A floor is not a substitute for compatible targeting, creative eligibility or a request to the intended unit. Change settings through the responsible owner's review with a preserved baseline.

Is campaign estimated cost money AdMob charged me?

Google defines it as estimated forgone revenue, not a campaign fee. Keep that modeled amount separate from publisher earnings, invoices and payments, and avoid treating it as measured loss.

Research and scope#

Reviewed 4 November 2026. The linked original questions are older evergreen reports; the opened pages expose relative ages rather than precise creation timestamps. Their account claims and suggested causes are not independently verified. Current official references support the platform distinctions; the workflow, checklist and arithmetic are diagnostic recommendations, not an executed app audit or an outage announcement.

Your next step

Check eligibility against the actual request

A 100% priority setting does not ensure delivery on every impression. Discuss what request and serving evidence is available.

Discuss a measurement question

A short inquiry to scope relevant options; no campaign access is needed.

Back to all field notes