Myth: the network filters out the fraud for you

Wrong, and naive. The network's IVT filter is tuned to its own liability, not your ROI. It catches the crude datacenter bots that would trigger an MRC complaint — and waves through the sophisticated stuff, because sophisticated invalid traffic still bills as a click, and that click is the network's revenue. They are structurally incentivized to under-detect what you're paying for.

Ask the obvious question: why would the party getting paid per click be aggressive about deleting clicks? Run your own post-click sniff test — session depth, time-to-bounce, IP entropy per placement. The placements with 90% sub-2-second bounces aren't bad luck. They're fraud the "filter" let through because catching it costs the network money.