Our in-house anti-fraud pipeline filters invalid traffic in under 50 milliseconds — before any advertiser is billed or any publisher is credited. Independent benchmarks put our detection rates on par with, and in some categories above, the leading third-party solutions.
An impression doesn't get billed until it survives every stage of this pipeline. Each stage catches what the previous one missed.
Known-bad IPs, datacenter ranges, suspicious user agents, and TOR/proxy exits are blocked before a single bid is placed.
240+ signals — IP reputation, device fingerprint, click velocity, viewport, behavioral anomalies — combined into a risk score in under 50ms.
Viewability, time-on-page, scroll depth, and conversion-path coherence checked after the impression to catch sophisticated fraud (SIVT).
Nightly cross-check against updated fraud signatures, MMP attribution windows, and conversion postback validity. Stragglers get clawed back.
Anything that slips through is refunded automatically within 30 days. No claim, no ticket — your balance just goes back up.
The IAB recognizes two classes of invalid traffic: GIVT (general — bots, crawlers, known patterns) and SIVT (sophisticated — adversarial, evolving). We catch both.
Traffic originating from AWS, GCP, Azure, OVH and other known cloud providers — the most common bot source.
Low-quality manual clicks from regions known for click-farm activity. Detected via velocity, device clustering, and conversion-rate patterns.
Apps that fake clicks immediately before install to steal attribution. Detected by timing anomalies between click and install events.
Bot traffic disguised as residential IPs via proxy services. Detected via TLS fingerprinting, behavioral patterns, and proxy-network signatures.
VPN, GPS spoofing, and proxy traffic claiming Tier-1 origins to harvest higher CPM. Cross-checked against device locale and timezone signals.
Ads stacked, hidden, served in 1×1 iframes, or rendered off-screen. MRC-standard viewability measurement on every impression.
Repeat installs from the same device with reset IDFA/AAID, often via emulators or factory-reset farms. Detected via device-fingerprint persistence.
Search engine bots, scrapers, and automation tools (Selenium, Puppeteer, headless Chrome). Detected via user-agent and behavioral fingerprints.
Fabricated install/conversion events from spoofed SDK calls. Detected via cryptographic SDK signatures and event-ordering validation.
Independent benchmarks across our 2025 production data. Vendors anonymized for competitive reasons. Methodology and raw data available on request to enterprise customers.
Why we publish this: Most ad networks treat fraud as a black box. We benchmark because our advertisers and publishers deserve to know they're getting verified traffic — without paying a 3-7% third-party fraud tax on top of their media spend.
Invalid impressions and clicks are filtered before billing. Anything that slips through is automatically refunded within 30 days. No claim forms, no waiting periods, no third-party fraud-detection tax of 3–7% on top of your media spend.
We filter at the inventory level too — so the eCPMs in your dashboard reflect real, monetizable traffic. Bot impressions don't depress your real eCPM, and they don't create chargeback risk later.
Member & TCF v2.2 compliant
Certified information security management
Privacy-compliant in 240+ GEOs
Standard-aligned measurement
Want to see methodology, raw data, or set up custom rules for your account? Email our security team →
Free to register. Anti-fraud included on every campaign and every publisher placement from day one.
Enterprise customer? Request the full methodology + raw data — happy to share under NDA.