1. Pre-bid and post-bid
Do you perform both pre-bid AND post-bid analysis, with a technically appropriate measurement after delivery? If not, why not and which parts of actual delivery remain unobserved?
2. Full measurement or sample
Is every technically measurable and successfully instrumented impression analyzed, or does the provider rely on samples and extrapolation? If samples are used, how is the risk of missing rare or deliberately hidden fraud patterns addressed?
3. JavaScript-based evidence
Is JavaScript-based measurement used where the environment allows it? If not, the provider should be able to explain which technical signals are lost and how that gap is compensated.
4. Which fraud type was identified?
An aggregate rate is not enough. Ask for a concrete explanation of which fraud or anomaly type was found, which signals support the finding and which alternative explanation has been excluded or remains open.
5. Protection against domain spoofing
How does the provider prevent allowlists or inclusion lists from being bypassed by domain spoofing, reseller chains or opaque supply paths? ads.txt alone is not a quality guarantee.
6. Complete delivery transparency
Do you receive domain and – where technically available – URL or placement, spend by placement, traffic quality, relevant bot/fraud types and positive evidence of human traffic?
7. Cross-signal view
Can platform, device type, operating system, browser, time, geography and other technical signals be examined together? Inconsistencies across signals can be more forensically relevant than a single score.
8. Audit process and cadence
How is the audit process structured, when is measurement performed, who interprets the findings, how often is it reviewed and how is it documented whether actions actually work?
9. Reporting with action context
Can reporting connect fraud categories, affected creatives, channels, timing, sources and proposed actions in a traceable way? An audit becomes valuable when a finding can be translated into a decision.
How to read the checklist
No answer is itself a signal. The point is not that every provider must use the same technical architecture, but that measurement coverage, limitations, evidence and decision logic can be explained transparently.
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.