How We Filter 98% of Bot Traffic Pre-Bid

Fraud protection only helps if it happens before you pay. Filtering after the fact means reconciling refunds and cleaning reports; filtering pre-bid means the bad impression never costs you anything.

Our Shield runs multiple checks on every request: device and network signals, behavioural patterns, and reputation data built from billions of daily impressions. Anything that looks non-human is dropped before a bid is ever placed.

Because the checks are layered, no single signal decides an outcome. A data-centre IP alone is not fraud, but a data-centre IP combined with impossible click timing and a spoofed user agent almost certainly is.

The result is that roughly 98% of non-human activity is removed pre-bid, so the traffic that reaches your campaign is overwhelmingly verified, real users.

Key takeaways

• Filtering happens before the bid, not after
• Multiple signals combine for each decision
• Reputation data spans billions of impressions
• ~98% of non-human activity removed pre-bid