Evidence
What audits actually find.
Every figure here comes from real cross-SSP spend, anonymised. We never name an advertiser, agency, deal or client.
Scale
Real budgets, reconciled to the cent.
These are real campaigns across multiple SSPs, not test data.
| Dataset | Impressions | Spend | SSPs | Days | Domains | Deals |
|---|---|---|---|---|---|---|
| Dataset A | ~136M | ~$1.44M | 3 | 263 | 17,664 | 86 |
| Dataset B | ~18.9M | ~$178k | n/a | 21 | 13,357 | 44 |
Authorisation
Roughly one route in seven couldn't be verified either way.
So we say so, rather than passing them as clean.
| Outcome | What it means | Routes | Share |
|---|---|---|---|
| Authorised direct | Confirmed by the publisher as a direct seller. | 2,564 | 40.0% |
| Authorised reseller | Confirmed by the publisher as an approved reseller. | 2,761 | 43.0% |
| Needs further review | Approved at the first step, with further steps still to confirm. Not treated as a violation. | 36 | 0.6% |
| Unverifiable | The evidence wasn't available to confirm it either way, so we report it separately. | 947 | 14.8% |
| Unauthorised | The publisher does not approve this seller for its inventory. | 109 | 1.7% |
On a separate 300-domain scope: 1,675 routes, 7 unauthorised, 274 unresolvable.
The rating gate
We declined to score a long tail we couldn't assess.
The rest were held as UNRATED for thin signal. The rated set carries the overwhelming majority of spend.
Cross-SSP honesty
All three SSPs were present on only 8.8% of the spend.
We state this next to the total, because a blended cross-SSP number over feeds that don't overlap isn't a like-for-like comparison.
Environment
Every page we rendered, and every one we couldn't.
| Outcome | Domains |
|---|---|
| Rendered successfully | 61 |
| Excluded as CTV / app identities | 31 |
| Errored | 5 |
| Bot-blocked | 3 |
Built-in safeguards
When the data looks wrong, the audit stops.
Checks run on every audit to catch bad inputs before they turn into findings.
Data integrity checks
On one dataset, a key field in the delivery data had been mislabelled across 683,660 rows. Taken at face value, it would have made almost every route look unauthorised. Our checks caught it, the audit paused, and nothing was reported until the data was corrected.
Consistency testing
Every view of the results is tested against the full analysis before it's shared. That testing once caught a view overstating spend threefold, long before any client saw it.
See what your own reports show.
Start with one campaign. We'll show you what checked out, what didn't, and where the money went.