Invalid traffic is not an abstract brand-safety problem. It is a line item on your ad invoice that was never going to convert. The goal of this piece is to give you a working vocabulary, the three categories of fraud that matter most to performance marketers, and a sense of what “good” looks like after you filter for them.
A working definition
Invalid traffic (IVT) is any click or impression generated by a non-human source, a misconfigured measurement source, or a deliberately fraudulent one. Industry bodies (MRC, IAB) split IVT into general (GIVT — known bots, datacenter traffic, crawlers) and sophisticated (SIVT — fraud rings, click farms, malware-driven activity, hijacked devices).
Both categories cost you money. The first is easy to filter on the platform side and largely already is. The second is where the bleed happens.
Three categories worth naming
These are the three patterns that drive the bulk of wasted spend for Shopify and Amazon operators.
- Click farms on display and discovery placements — high CTR, near-zero add-to-cart rate, near-zero LTV. Often concentrated on a small set of low-quality placements.
- Bot-driven retargeting loops — sessions that bounce, get re-served, click again, and bounce again. The ROAS report looks fine; nothing reaches checkout.
- Misattributed organic — not fraud in the criminal sense, but pixel events fired by browser extensions, accelerated mobile pages, or proxy traffic that get credited to a paid source they did not originate from.
What good looks like after you filter
Most operators see 8–18% of declared paid traffic flagged invalid in their first two weeks of monitoring. The headline number is not the point — the point is the placement-level distribution, because that is what unlocks exclusion lists you can actually act on.
After fraud filters run for a few weeks, your POAS by placement starts to look very different. Cheap placements that were carrying suspicious volumes drop out of the top of the queue; trusted placements move up. The cleaner POAS is itself the proof the filter is working.
- IVT is a spend leak, not a brand-safety abstraction.
- The three patterns worth monitoring are click-farm placements, bot retargeting loops, and misattributed organic.
- After 2–3 weeks of filtering, the right diagnostic is whether POAS by placement has visibly redistributed — not whether the invalid-traffic percentage went down.