Shared negative lists: account-wide or per campaign once you have several campaigns?

Once an account runs more than a few campaigns, I'd put the shared list to one job only: true noise. That means words like 'jobs', 'free' and 'reviews of', which are junk everywhere, so nobody has to maintain them in five places.

The trouble starts when a term is junk for one service but valuable for another. Say a dental clinic blocks 'cheap' account-wide. Now the implant campaign never sees 'cheap implants' searches, and those can come from people who actually book.

Campaign-level lists take more upkeep, but you control each one. So my approach is a small shared list for real noise only, with anything that carries commercial meaning pushed down to the campaign it belongs to. How do you handle the split?

A shared negative list applies everything as broad match negatives, so "cheap" also blocks "implants cheap" and "cheap dental clinic", not just the exact phrase. Attach that list to every campaign and one word can quietly kill a service line. For terms that are only junk in one context, add them at ad group level or as exact match instead, then watch the search terms report for a couple of weeks to see what they're actually blocking.

The ad group fix only works for Search and Shopping. PMax has no ad groups, so its negatives sit at campaign level or in a shared list, which means a "cheap" in a shared list hits PMax too, and nothing in the interface warns you when it does. Exact match isn't a clean escape either, because it misses variants you didn't think of, like "cheaper implants" or a typo. For a commercial phrase that only matters in one context, a multi-word phrase match such as "cheap implants" usually fits better. It blocks that sequence and leaves "implants cheap" alone. A single word in phrase match behaves almost like broad, so the trick only works with two or more words. Are you running PMax alongside Search in the same account?

The two-word phrase trick is useful, and for local services I'd take it further. A plumber or clinic rarely wants a commercial word blocked everywhere, so I'd keep those words out of shared lists entirely and add them as campaign negatives only after the search terms data shows they're actually junk. PMax makes that harder because it hides a lot of queries, so a negative can keep blocking searches you never see. Does the PMax search terms insight show you what a campaign-level negative is blocking?

I'd be careful about assuming the insight shows you blocked queries. As far as I know it doesn't. A negative stops the ad from showing, so a blocked query never gets an impression and never reaches the search terms data. You only see what got through.

So the check has to run the other way. Before adding a campaign negative, test the phrase in a Search campaign without that negative, or watch conversions for that service line for a couple of weeks after you add it. A drop in booked calls is the signal that the negative is hurting you.

Agreed on the blind spot, but I'd count the watch window in conversions, not days. Two weeks only works if the line books a lot. A clinic getting around 8 implant bookings a month will see a two-week dip that's mostly noise, so I'd wait for 20 to 30 conversions on that line before judging the negative.

There's also a cheaper check before you add anything. Pull the last 90 days of search terms and look for that phrase. If any of those queries already converted, the phrase isn't pure junk, and a campaign negative is the wrong tool. That quick filter often settles the question before you touch the list.

How many bookings does your implant line get in a typical month?

I'd put a spend threshold on that filter. Zero conversions only means something once a phrase has burned through roughly 2 to 3 times your target CPA. Before that you're mostly looking at normal variance. If a phrase has spent $60 against a $25 CPA with nothing booked, that's a real negative candidate. One that's spent $12 just hasn't given you enough data yet. Combine that with your converted-query check and the rule stays simple: only block a phrase once it's clearly past that spend line and nothing has ever converted.

The 2 to 3x CPA line works, but I'd set it per service line instead of per account. A $25 target might make sense for cleanings, while an implant booking can run several hundred dollars. One flat figure would either block implant phrases too early or let cleaning junk run far too long.

So the rule becomes: a phrase gets cut once it has spent 2 to 3x the target CPA of the line it sits in, with zero conversions and no converted query in the last 90 days. Anything short of that stays in the data a while longer.