The starting point: aggregation hid anomalies
Fou describes campaigns from his agency years with placement-level CTRs above 100% and even domains designed to resemble legitimate medical information sites. Aggregated campaign averages made those anomalies much less visible.
Why conventional web analytics were insufficient
Landing-page reports showed high bounce rates and little interaction but did not explain who or what generated the visits. That led Fou to build a JavaScript measurement tag with greater forensic detail.
300–500 parameters, no cookies, no PII
Fou says the platform collects hundreds of browser and interaction parameters deterministically. Cookies and personally identifiable information are not required for the technical consistency and automation analysis he describes.
Combine red flags and blue flags
Datacenter IPs, implausible user-agent and screen combinations or missing browser properties can become red flags. Mouse, scroll, touch and other plausible interaction signals can become blue flags. No single signal is treated as sufficient proof.
Detections evolve with the attack pattern
Fou emphasizes that detection logic has been tuned over years as fraud methods changed. A static rulebook is therefore not a durable solution.
Cybersecurity and interpretation are part of measurement
The platform is designed to make tampering and false data injection harder. Fou also stresses practitioner analysis so that technical signals are interpreted in campaign and business context.
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.