How Advlume Detects Fraud

The layered fraud detection stack: real-time ADMD client-side scoring (blocks score-1 sessions immediately), Kueez RTB signals, and Google Ad Manager's overnight server-side reconciliation.

Advlume uses a layered approach to detect and block invalid traffic before it can harm your account or your demand partners' campaigns. Detection happens at two levels: real-time client-side scoring and server-side reconciliation by Google Ad Manager.

Client-side: ADMD traffic quality gate

Every page session that loads the Advlume wrapper is scored in real time by ADMD (a ClearTrust-based traffic quality module). The monitoring script is injected asynchronously after the wrapper loads, so it does not slow page rendering.

The scoring process:

  1. The monitoring script collects browser signals (user-agent, screen properties, timing patterns, mouse/touch activity, headless browser indicators) and sends them to a scoring endpoint.
  2. A score between 0 (clean) and 1 (high risk) is returned.
  3. The wrapper passes this score to the GAM display gate:
    • Score 0: Normal ad serving, no restriction.
    • Score 0.1–0.9: Ads are served. The session is flagged and the score is reported to the real-time dashboard. Demand partners see this signal via bid request enrichment.
    • Score 1: All ad rendering is immediately blocked. GPT never refreshes. No impressions are recorded for this session.
  4. The score is also emitted to the real-time ingest endpoint so your Live dashboard shows the blocked session count for each site.

Kueez RTB anti-fraud layer

The Kueez RTB demand partner SDK (static.kueezrtb.com/latest.js), loaded alongside the wrapper, includes its own anti-fraud scoring. It contributes additional signals to the server-side reconciliation used by Kueez's own demand.

Server-side: Google Ad Manager reconciliation

GAM applies post-delivery IVT filtering overnight. After each day closes, GAM's internal system reviews all recorded impressions and removes those attributed to known invalid traffic patterns — data centre IPs, known bot signatures, and statistical anomalies. Impressions removed during this process are deducted from your revenue figures, which is why prior-day revenue numbers occasionally decrease the following morning.

Demand partner filtering

Each DSP connected to your inventory runs its own independent fraud detection. A DSP that detects high IVT signals on a given domain will suppress its bids for that domain — this shows up as reduced bid density and lower eCPMs in your reporting. Advlume's client-side ADMD gate significantly reduces the IVT signals that reach DSPs, which is why average CPMs on Advlume-monitored inventory typically improve over time.

What you can do

Detection is automatic and requires no configuration on your part. If your Live dashboard consistently shows a high percentage of blocked sessions (typically >5% is worth investigating), see Avoiding IVT on Your Site for guidance on traffic sources to audit.

Further reading

Last updated 2 months ago