Skip to content
All field notes
Measurement
Published
Oct 30, 2026
Updated
Nov 4, 2026
Reading time
13 min read

AdMob vs Firebase Ad Revenue: Find the Reporting Gap

Audit automatic and manual ad-revenue collection, callback units, precision and GA4 report scope before fixing an AdMob versus Firebase revenue discrepancy.

AP
AdProof Research
Independent measurement desk
One mobile ad sends a revenue receipt along two separate collection paths toward different report sheets, with a duplicate copy held for inspection.
Automatic collection, custom revenue events and reported ad revenue need separate evidence. Conceptual illustration without customer figures.
In this guide (11 sections)

When AdMob and Firebase ad revenue disagree, first identify who produced each revenue record and what each report sums. An automatic publisher integration, a manual paid-event callback, a custom event value and GA4's Total ad revenue are different checkpoints. Trace one supported collection path before adding another event or changing the value to make totals agree.

A real Android developer question about automatic and custom ad-revenue events describes a custom OnPaidEventListener event exceeding the automatic result. Two comments report similar symptoms, but the discussion does not establish a cause. Its claimed difference is not an acceptable-discrepancy benchmark.

An older AdMob discussion recommending manual revenue events also contains the useful question: why add a manual event when an automatic one already exists? The answer depends on the actual monetization and collection arrangement. “Always log it again” is not a reliable integration rule.

Define the comparison before inspecting code#

Save the two original reports, not just their headline totals. Record the app, platform, app versions, ad sources, ad units, formats, dates, timezone, currency, extraction time, exact metric and any filters.

Then distinguish these records:

RecordUseful questionKeep separate
AdMob reportWhat revenue does this named report include?Bank payment and custom event sums
Automatic Analytics integrationWhich linked app and supported path supplies data?Manually logged copies
SDK paid-event callbackWhat value, unit and precision did this ad object supply?The receiver's reporting result
Manually logged eventWhich code path sent it, with which parameters?Automatic publisher revenue
GA4 reportWhich metric and population does the selected view summarize?A different event's parameter total

Google's Monetization overview reference distinguishes Total ad revenue from Total revenue, which also includes other revenue categories and refunds. Comparing the broader total with an ad-only source creates a definition gap before any collection defect is considered.

Use a completed, consistently scoped interval. A broad Analytics property can include streams or versions outside the AdMob export. A custom event captured only by the newest release can cover less inventory even when its code works correctly.

Related guide

Trace the ad revenue collection path

Record each revenue source and its report before comparing totals.

Use the tracking QA checklist

Opens a practical guide. No email required.

Establish one collection owner for the intended revenue#

Firebase's ad-revenue measurement guide describes automatic ad_impression collection when the AdMob app is linked to Firebase and Analytics. For other monetization platforms, it describes using their impression-level revenue data to log the event.

Draw the route actually installed:

Ad impression → rendering SDK or mediation owner
              → automatic integration OR documented manual collection
              → intended Analytics app/property
              → selected reporting surface

This is an ownership diagram, not a recommendation to toggle live links during diagnosis. Identify every additional route: a wrapper callback, another analytics SDK, a custom event, a server relay or a callback registered again on reload.

Ask which route supplies the number the team calls revenue. A separate diagnostic event does not necessarily inflate a native revenue metric simply because it exists. Inspect the event name, parameters, report definition and any downstream sum. Conversely, an apparently convenient combined chart can add the same underlying value from two routes if its query does so.

Keep intended diagnostic observations outside the production revenue calculation unless their purpose and treatment are explicit. Do not log invented revenue to validate a live dashboard.

Treat the third-party mediation duplication warning precisely#

Google's AdMob data-flow guide specifically warns that when a third-party mediation platform supplies impression-level data for manually logged Analytics ad_impression events, also linking AdMob to Firebase can duplicate revenue data.

That warning is scoped to the described arrangement. It is not advice for every AdMob publisher to unlink Firebase. First establish who mediates the impression, which networks the manual callback includes, and what the automatic link contributes. Review the supported integration with its owner before changing collection.

The same guide distinguishes real-time impression-level data from reporting adjustments and describes refreshes within 72 hours for specified Analytics revenue flows. It also states that reported Meta Audience Network revenue remains estimated and points to that network's own revenue as definitive for its earnings. Preserve source and observation time instead of expecting every callback sum to equal every later report.

The same guide distinguishes real-time impression-level data from reporting adjustments and describes refreshes within 72 hours for specified Analytics revenue flows. It also states that reported Meta Audience Network revenue remains estimated and points to that network's own revenue as definitive for its earnings. Preserve source and observation time instead of expecting every callback sum to equal every later report. For a LevelPlay-versus-Meta reporting question, identify the report family and its valuation basis before treating a pivot prediction or loaded-ad value as network earnings.

Do not convert that processing description into a universal “wait 72 hours” repair. Waiting cannot correct the wrong destination, a double submission or a unit error you have already reproduced.

Inspect the actual callback and its units#

Record the released SDK family and version, rendering adapter, ad format, ad object and the code responsible for registering the paid-event listener. Current documentation can describe a different SDK branch from the app installed on a device. If the disagreement concerns daily app-user totals rather than callback revenue, compare Play, Analytics and RevenueCat observation contracts before attributing it to missing ad-revenue collection.

The Android Legacy impression-level revenue guide recommends registering the listener as soon as the ad object is available and before showing it, then forwarding the revenue observation promptly. A listener installed after the relevant callback can leave a collection gap. Confirm the behavior appropriate to your actual SDK rather than copying method names between platforms.

Google's AdValue reference defines micro-units: 1,000,000 micro-units equal one unit of the reported currency. Preserve the original integer and currency alongside any normalized diagnostic value.

Hypothetical unit check: a callback value of 5,000 micros is 0.005 currency units, not 5,000 units and not five units. For an implementation using integer arithmetic, check that division does not discard the fractional value before it is logged. Avoid rounding every small impression value to cents before aggregation; retain suitable precision through the chosen reporting pipeline.

This unit check is not a production event implementation. Verify the destination's parameter contract, numeric type and currency requirements before changing a released integration.

Preserve precision instead of relabeling every value exact#

Google's impression-level revenue explanation identifies these precision categories:

CategoryMeaning to retain in the investigation
PreciseThe supplied precise value for the ad under the documented source coverage
EstimatedA value estimated from aggregated data
Publisher providedA configured manual CPM basis for the relevant waterfall source instance
UnknownInsufficient available information; preserve the category rather than inventing a price

Record the category with the value. A configured manual CPM does not establish an impression's actual auction clearing price or a finalized bank receivable. Split comparisons by source, format and precision where those fields are available. Keep missing categories visible instead of dropping them to improve agreement.

The same reference calls for enabling impression-level revenue when using the documented Firebase export flow. Verify that setting and the platform prerequisites with the implementation owner; a current dashboard configuration does not prove an older installed app contains the required callback path.

Do not multiply a callback value by a guessed revenue share or change its precision label because another report looks higher. An unexplained discrepancy needs evidence about its population and processing.

Check the reporting surface before calling a missing parameter a defect#

The automatically collected events reference says the default publisher ad_impression integration does not pass value or currency, and that this automatic event is not exported to BigQuery. That is a different data path from a custom event whose own parameters you export and sum.

A missing value on that automatic event therefore does not, by itself, establish missing aggregate publisher revenue. Nor should a BigQuery query be expected to reproduce an automatic event stream the documentation says is not exported. Preserve whether your evidence comes from an automatic integration, a manually logged event or an aggregate report.

Use the Publisher ads report reference to name the selected dimensions and revenue metric. A custom paid_ad_impression event name alone does not establish that its parameter total is the same measure as Total ad revenue. Retain the exact query or report configuration that produced the custom total.

If Measurement Protocol is involved, Google's event reference adds another boundary: an ad_impression sent that way is not exported to advertising platforms such as Google Ads. It also warns against sending a copy already captured by tagging or the SDK. A server replay is not a universal way to repair an advertising export.

Test a bounded path, with development traffic labeled#

Use the documented test-ad setup appropriate to the actual SDK and each mediation network. Confirm test status before interacting; do not generate production clicks to obtain a useful revenue value.

In a controlled build, capture this sequence:

  1. Which ad object and supported rendering path produced the observation.
  2. When the paid-event listener became active.
  3. The callback's original value, currency and precision.
  4. Each deliberate forwarding attempt and its destination.
  5. The corresponding event and parameters exposed by receiving diagnostics.
  6. The report definition and later observation used for comparison.

The Android Legacy guide documents UNKNOWN precision and zero ad value for bidding-source test requests. That result can be expected under the described test conditions. It does not establish that production paid revenue is zero. Use the test to examine wiring and ownership, then keep it separate from commercial reporting.

Firebase DebugView supports development-device event and parameter inspection. It can help verify the event you actually log; it does not establish finalized revenue or complete production coverage. Record debug and test status explicitly, and confirm their current report/export treatment rather than assuming development activity is automatically absent everywhere.

The tracking QA checklist provides a broader release and receiving-evidence record. Rerun the affected route after an authorized change using the same versioned test conditions.

Worked example: separate two demonstrated defects from the remainder#

Hypothetical teaching example. Assume an authorized diagnostic ledger identifies 1,000 comparable callback observations in one currency. It contains 5,000,000 micros, equal to 5.00 units. The comparison below concerns a separately defined custom-event ledger, not a claim that automatic publisher events are exported individually.

Suppose the custom ledger sums to 6.20 units. Controlled evidence then establishes:

ComponentUnitsEvidence required
Comparable callback value5.00Original values, consistent currency and defined scope
Repeated forwarding0.50Matching observations sent twice by a reproduced retry path
Events outside the selected app/version scope0.40Verified population mismatch
Still unexplained0.30No demonstrated allocation or cause
Custom-ledger total6.205.00 + 0.50 + 0.40 + 0.30

Removing the verified repeats and scope mismatch from the comparison leaves 5.30 units, a 0.30 unresolved difference. It does not prove an extra 0.30 of earned money, lost impressions, invalid traffic or billing recovery.

If a later aggregate report shows another amount, compare its definitions and maturity separately. Do not force its total into this callback ledger or invent a second duplication diagnosis from the arithmetic alone.

Choose the next action from the first missing boundary#

FindingNext useful action
Automatic and manual routes cover the same intended revenueReview the supported collection arrangement and prove overlap before changing it
Original callback and forwarded amount use different unitsCorrect and retest the documented conversion in an authorized build
One format or released version lacks callbacksInvestigate its listener lifecycle and rendering path
A custom query includes other apps or event producersCorrect the comparison scope while preserving the original export
Automatic event lacks an individually exported valueUse its documented aggregate reporting path; do not infer zero revenue
Comparable values still differRetain the unexplained portion with source, precision and processing limitations

If your callback count comes from rewarded-ad server-side verification rather than an impression-level revenue listener, reconcile verified reward events against the selected AdMob report. Reward notifications, impression observations and revenue values require different counting rules.

For a difference between earnings and cash received, use the separate AdMob finalized earnings and payment reconciliation guide. Callback collection does not establish a settlement or reveal a private deduction reason. If the confusing field is a house campaign’s estimated cost, separate campaign delivery from modeled forgone revenue before treating it as a fee or earnings shortfall.

For campaigns already measured with AdProof, retain a supported campaign-event or aggregate export with matching dates, timezone and filters alongside the independently obtained app evidence. The anonymized FMCG count-once case illustrates delivery-count reconciliation; it does not demonstrate a Firebase integration or explain publisher payout. Discuss the available campaign evidence before assuming a particular app revenue path is independently covered.

Frequently asked questions#

Should I manually log revenue when AdMob is already linked to Firebase?

First identify the collection arrangement and report you need. Automatic AdMob collection and documented third-party monetization paths differ. Do not add another production route without checking its overlap and supported treatment.

Does PRECISE mean the callback must equal my eventual payment?

It describes the supplied ad value's precision. A payment involves a different period, population and settlement record. Preserve the precision category and reconcile financial statements separately.

Why is value absent from the automatic ad_impression event?

Google documents that behavior for the default publisher integration. Inspect the supported aggregate revenue report rather than treating a missing event parameter as proof of zero earnings.

Does a zero value in a bidding-source test prove my integration is broken?

No. The cited Android Legacy guidance documents zero value with UNKNOWN precision for that test case. Verify the callback and destination behavior under the appropriate test conditions, then assess production reporting separately.

Research and scope#

Reviewed 30 October 2026. The original questions are older evergreen reports. They establish an actual implementation problem, not a current platform outage or an acceptable error percentage. Official references were opened during research; platform and SDK-family limits remain explicit. The workflow and arithmetic are diagnostic recommendations, not an executed app audit or a revenue guarantee.

Your next step

Trace the ad revenue collection path

Record each revenue source and its report before comparing totals.

Use the tracking QA checklist

Opens a practical guide. No email required.

Back to all field notes