The starting point: two reports and an unresolved question
Fou describes a campaign reported as virtually fraud-free and 99.84% CTV by verification. The cited DSP breakdown instead shows 32.9% Connected TV, 49.8% PC, 16.5% mobile and 0.8% tablet. According to his account, other devices had not been excluded in the setup. These reported figures warrant investigation; they are not an audit finding independently confirmed by H.O.T. AG. [1]
First establish the denominators
A percentage describes a numerator and a denominator. Before comparing reports, campaign IDs, period, time zone, filters and event definitions must align. If one report covers all billed impressions and another only classifiable events, they answer different questions. Spend shares and impression shares are not interchangeable either. An enabled device option establishes eligible inventory; it does not measure the actual delivery distribution.
Three separate requirements for evidence
First: did delivery meet the agreed device and inventory scope? Second: which events could be assessed, and by which method? Third: which specific observations support an IVT or fraud classification? Answering one question does not resolve the others. A valid human contact on a PC may breach an explicitly agreed TV-screen buy without being bot traffic.
Why the DSP report also needs scrutiny
The DSP breakdown is an important comparison point, but not independent proof of the receiving device. IAB Tech Lab explains that device information in bid requests and measurement beacons can be manipulated. We therefore examine signal origin and transformation, available app and player information, and contradictory attributes. Agreement between reports adds evidence only when they are not simply repeating the same unverified input. [3]
CTV requires a disclosed measurement architecture
With server-side ad insertion (SSAI), events may be reported by a server on the player’s behalf. Such a request is not automatically a direct observation on the television. IAB identifies platform access and SSAI as verification challenges. The report must therefore disclose where a signal was collected and which implementation supports the conclusion drawn from it. [2]
Unknown must remain visible
MRC guidance requires unknown rather than valid treatment for certain SSAI situations that cannot be validated. Conversely, data-centre or proxy origin alone does not establish IVT. We expect separate quantities for captured events, technically assessable events and unresolved cases. An overall percentage must not conceal these distinctions. [4]
How we would investigate the discrepancy
We preserve the original exports and configurations, align scope and reconcile available event identifiers. Next, we assess device classification and measurement integration using accessible technical evidence. Suspect sources are documented individually, with alternative explanations retained in the findings. FouAnalytics can contribute additional quality signals where the environment actually exposes them. The result separates confirmed misplacements, classification problems and cases that remain undecidable.
The implication for buying and governance
Contracts should define the intended meaning of CTV, permitted devices, measurement methods, data access and escalation deadlines. Credits or exclusions require defensible attribution. After a correction, the same scope is measured again. H.O.T. AG supports advertisers through media consulting or a transparently separated audit role. The decisive question is: which observation and data basis support this percentage?
Sources and technical references
The campaign figures come from Fou’s original article. H.O.T. AG has not independently examined the raw data or the unnamed verification vendor’s implementation. Potential causes are investigation hypotheses. No general error or fraud rates are inferred from this individual case. Statements about FouAnalytics depend on the applicable technical measurement scope.