Which traffic sources to cut first
You have a list of sources and a bot percentage next to each. Cutting them in the wrong order costs more than not cutting at all.
Last updated
Rank by wasted spend, not by bot share
A zone at 90 percent bots that sent 40 visits wasted almost nothing. A zone at 55 percent that sent 40,000 visits is where your money went. Sort by the number of bot visits, not by the percentage, and the list usually reorders completely.
Cut confirmed before suspected
A confirmed bot was verified: a named crawler, a data center address, or a client that failed a check it claimed to pass. A suspected bot scored above a threshold and nothing more. Cut the sources whose traffic is mostly confirmed first, because those decisions are the ones you can defend if the network pushes back.
Use the trend before the level
A source that has always run at 30 percent bots is a known cost you have been paying and pricing around. A source that went from 5 percent to 30 percent this week changed, and something caused it. The quality score carries a trend for exactly this reason, and the trend is usually the better signal for what to cut today.
Cut at the narrowest address that works
If a single placement inside a publisher is bad, block the placement rather than the publisher. Blocking too wide is invisible: you never see the good traffic you stopped buying, so the mistake never announces itself. Export one list per field so the narrow option stays available to you.
Keep the receipts
Before you cut, download the evidence CSV for that range. Once you block a source it stops sending traffic, and the record of why you blocked it stops accumulating with it. If you plan to ask for a refund later, that snapshot is the argument.