Discrepancy Investigation — Template

Field	Details
Campaign	Campaign name
Advertiser
IO Number	IO-XXXX
Discrepancy Type	Impressions
System A	Your system (ad server, SSP, exchange, publisher)
System B	Their system (third-party, DSP, buyer, agency, exchange)
Investigated By	Your name
Status	Investigating

Metric

System A

System B

Difference

Diff %

Detail	Our Side	Their Side
Reporting period	YYYY-MM-DD — YYYY-MM-DD	YYYY-MM-DD — YYYY-MM-DD
Counting method	Ad server impression = served / rendered / viewable?
Data source	GAM, Xandr, Freewheel — report name	DCM, Sizmek, MOAT — report name
Filters applied	Any geo, device, or line item filters

Work through these in order. Most discrepancies are caused by the first three items. Stop when you find the cause.

Confirm both reports cover the exact same dates
Confirm timezone alignment (UTC vs. Eastern vs. Pacific matters — a midnight boundary difference can shift a day's worth of impressions)
Check for partial-day data — did the campaign start or end mid-day?

Our ad server counts on serve vs. render vs. viewable — which one?
Their system counts on serve vs. render vs. viewable — which one?
If different: the gap IS the methodology difference, not a bug. Document and close.
Check if either system deduplicates (e.g., one user = one impression per session vs. per request)

Is the third-party pixel firing on every impression? Check with browser dev tools
Is the pixel inside the ad creative or on the page? (Creative-level fires more reliably)
Is the pixel blocked by CORS, iFrame sandboxing, or Content Security Policy?
Is the pixel loaded synchronously or asynchronously? Async can miss fast-scrolling users
Check for cached pages — cached impressions fire our ad server but may not fire their pixel

Ad blockers block third-party pixels but may not block our first-party ad server
IVT (Invalid Traffic) filtration — does one system filter bots more aggressively?
Check GIVT and SIVT rates in your verification vendor dashboard

What caused the discrepancy? Be specific — this is what you'll send to the buyer.

Finding	Details
Root cause	What specifically caused the number difference
Explainable gap	The portion of the discrepancy explained by this cause
Unexplained gap	Any remaining discrepancy after the root cause is accounted for
Within IAB tolerance?	IAB guidelines allow up to 10% discrepancy for display, 5% for video

Dimension	Details
Agreed action	Bill on publisher numbers / bill on third-party numbers / split the difference / makegood
Credit / adjustment	$X,XXX credit applied (if any)
Makegood	X,XXX additional impressions to be delivered by [date]
Prevention	What we'll change to reduce this discrepancy in the future

What should change so this doesn't happen again (or so it's caught earlier)?

Fix or action to prevent recurrence
Monitoring to add: automated discrepancy alert if gap > X% after day 2
Process change: reconcile third-party numbers at day 3 of every campaign, not end of flight

Track every touchpoint with the buyer. If this goes to finance or legal, the paper trail matters.

Discrepancy Investigation Template