Illustrative data: the 6,000-impression difference is 6% of the publisher count. The chart shows a discrepancy, not its cause.
An ad server recording an ad response and a third party recording a rendered impression are observing different points in the delivery path. Redirects, abandoned pages, blocked requests, and filtering can create a difference without either system having a simple counting bug.
A gap around midnight is a reason to inspect timezone boundaries, not proof of a timezone error. There is no universal acceptable discrepancy percentage for every platform and contract. Use the agreed measurement basis and a comparable historical baseline.
Calculate the discrepancy with a named baseline
Use (reference count minus comparison count) divided by reference count, multiplied by 100. Name the reference system because changing the denominator changes the percentage.
Reconcile the reports before editing tags
- Match campaign, placement, creative, date range, and timezone.
- Confirm whether each count means served, rendered, viewable, or another event.
- Compare finalized data with finalized data and note filtering or deduplication.
- Break the difference down by day, device, creative, and inventory source.
- Where IDs are available in both systems, inspect matched events and tracking responses. Otherwise compare a controlled test and aligned aggregates.
Choose a fix that matches the evidence
If the gap disappears after aligning the reports, document the reporting convention. If a specific creative never fires a tracker, reproduce and fix that creative. If the divergence appears after filtering, investigate traffic quality and processing rules.
After the correction, rerun the same comparison. Share the counts, formula, scope, evidence, and remaining uncertainty with both reporting owners. Do not describe an aggregate difference as proven missing impressions without tracing the event path.
Test your reasoning: Choose the denominator
Original simulated exercise. This is not a real client case.
Two aligned reports show different counts.
Evidence available
- Publisher: 80,000 rendered impressions.
- Third party: 76,000 comparable impressions.
- Both use the same finalized window.
What is the shortfall relative to the publisher?
Read the answer and next check
The shortfall is 4,000 / 80,000 x 100 = 5%. This measures a difference but does not identify its cause.
Next check: Segment the gap by creative and device, then inspect the event path where it concentrates.