The logic gap behind “99% good”
If a verification system reports 1% IVT, it does not logically follow that the other 99% is quality. It means only that the remainder was not classified as invalid by those rules.
Monitored is not the same as measured
Fou highlights the difference between the total number of ads monitored in reporting and the subset of impressions actually measured with a tag. Unmeasured delivery should not silently become confirmed quality.
Positive quality is its own evidence question
Quality may require additional evidence such as a plausible device and browser context, human interaction, consistent delivery data, a real site or app environment and traceable supply paths.
Aggregation can hide uncertainty
A single IVT rate can obscure differences across publishers, apps, placements, devices or time periods. A robust audit therefore examines coverage and distribution, not only the average.
Unknown is a valid category
Where observation is incomplete, “unknown” can be more rigorous than a forced classification. Governance should distinguish confirmed quality, confirmed problems and insufficiently observable areas.
This article describes principles and questions. Specific fraud, bot or quality rates cannot be transferred directly to individual campaigns without understanding methodology, scope and data basis.