Understanding ad serving discrepancies between systems

An ad-serving discrepancy is a difference between systems’ counts for apparently comparable delivery. First check whether the reports cover the same placements, period and counting event. A publisher’s selected creative and an advertiser’s recorded impression need not be the same event.

Counts can differ because an impression may be triggered when an ad loads or only when a portion is visible. A served impression and a viewable impression are not interchangeable; viewability is a stricter count.

Align the reports

Match placement, creative or tag, format, date range and time zone. Check whether either report includes defaults, companions, tests or other placements. Record each report’s extraction time; late data and filtering can change totals.

In Google Ad Manager and Campaign Manager 360, select reports that can be filtered to the same campaign, placement and date range. Record each report name and the exact metric label selected; similar labels do not prove that the reports count the same event.

Calculate the gap with an explicit denominator. For example, (publisher count − advertiser count) ÷ publisher count uses the publisher total as its base. A percentage based on the advertiser total answers a different question.

For each lead, keep the campaign and placement filters matched and note the metric label used in each system. Compare the counts at the same date range, then use the delivery checks below to identify the first point where they stop matching.

These are leads, not diagnoses. A gap’s direction does not identify its cause.

Find the counting stage where totals separate

Google Ad Manager and Campaign Manager 360 each document how impressions and clicks are counted. Their methods are not necessarily identical, so similar column names do not guarantee identical events.

Treat an ad request, served impression, rendered impression and viewable impression as separate checkpoints, not synonyms. An impression may be counted at load or only when part of the ad is visible, so a served count can differ from a viewable count.

For a browser-based third-party display tag, check publisher selection, the advertiser-server call, asset response and expected impression requests in order. Compare each hand-off and note the first point where the tag or its expected request is missing or changed.

Keep the final tag the publisher entered; a changed or wrapped export can behave differently from the original. Check cache busting only where the tag requires it. Use the provider’s macro documentation to confirm the required placeholder and escaping rules, rather than guessing.

Ad Serving Counting Events: Key Differences Between Systems

Record the resolution

Classify the difference as a population mismatch, counting-rule difference, tag defect, delivery failure or unresolved gap. Fix a confirmed implementation error, then compare a later defined period. If the methods differ legitimately, retain both definitions and record which metric governs the agreement. Do not apply a blanket correction percentage.

When investigating, check for changes or duplication, date mismatches, cache-busting setup, filtering and differences in measurement methods. To escalate, keep the relevant reports, test details and final tag available while the campaign can still be inspected.

Common Causes of Ad Serving Discrepancies