What impression discrepancy is acceptable?
Ten percent is the number everyone quotes. It comes from a contract template, it describes when parties must reconcile rather than when a gap is harmless, and it says nothing about where the impressions went.
Pixtru guides · updated
Where the 10% comes from
The 4A's/IAB Standard Terms and Conditions for Internet Advertising are the default contract terms for many US media buys. They define a Controlling Measurement: when both sides track delivery, the count used for invoicing, taken from an ad server certified against the industry's measurement guidelines.
If the other side's count differs from the Controlling Measurement by more than 10% over the invoice period, and the Controlling Measurement is the lower one, the parties undertake to reconcile. If a good-faith reconciliation fails, the terms give the agency a choice: treat the gap as under-delivery, or pay on the Controlling Measurement plus a 10% upward adjustment.
So the 10% is a trigger for a conversation written into a template. It is not a finding that a gap below it is correct.
And the 20%
Google's help pages for Campaign Manager 360 and Display & Video 360 call variances of up to 20% normal between their counts and a third party's. That is a support threshold: the point below which Google will not treat a gap as a fault in its systems. See DV360 discrepancies.
What "acceptable" costs
| Monthly media spend | A 9% gap | Over a year |
|---|---|---|
| $50,000 | $4,500 | $54,000 |
| $250,000 | $22,500 | $270,000 |
| $1,000,000 | $90,000 | $1,080,000 |
A gap inside the threshold is still impressions that were paid for and not delivered in any form a second system could see. The discrepancy calculator works it out for your own numbers.
A steady gap is a structural one
Random noise moves around. A gap that is the same size month after month has a cause: a creative trafficked without its tag, an exchange with slow pages, a share of readers whose content blockers stop the measurement. Structural gaps can be found, and many can be fixed. See when the tag never fires.
Setting your own threshold
- Name the Controlling Measurement in the insertion order, and which count it is — billed, served, rendered or viewable. See served, measurable, viewable.
- Agree the comparison: time zone, date range, and which invalid-traffic filtering applies.
- Ask for the gap explained, not just sized. "12% over" starts an argument; "7% never served, 5% served to a creative with no tag" ends one.
From acceptable to explained
The question worth asking is not whether a gap is under 10% but what is in it. Pixtru answers that per impression: each billed impression is given one reason at the stage it stopped, and the reasons add back up to the invoice. See why the numbers do not match.