Negative keywords from search terms: phrase match or exact match?

The exact vs phrase negative debate keeps coming up. Phrase is the safer block because negating "cheap plumber" also stops "cheap plumber near me". The catch is that it can take out variants you wanted to keep. In that same example, "cheap emergency plumber" wouldn't be caught, since the words aren't contiguous, but anything that does contain the phrase goes with it. Exact negatives are more precise, but the same waste keeps slipping through in slightly different wording.

My view: use phrase for clearly irrelevant intent (free, jobs, DIY, courses), and go exact only when a term is relevant in a different sense and you want to keep its close variants. Where do you draw the line when a term is only half-useful?

Phrase match negatives care about word order, so "cheap plumber" as a negative won't catch "plumber cheap" or "plumber free". For terms that are only half useful, I'd negate the modifier instead of the head term. Add "free" or "DIY" as single-word phrase negatives and leave "plumber" alone, so the good queries still get through.

Negating the modifier is the right instinct, but the same word often turns up in queries you want. "Free" is the obvious one: "free estimate" from someone ready to book gets blocked just as hard as "free DIY guide". Before adding a modifier as a negative, pull the converting terms that contain it. If some of them convert, a narrower phrase like "free DIY guide" catches the waste without touching "free estimate". Are you checking converting terms before you block a modifier, or only the wasted ones?

I'm not sold on the narrower phrase idea. "Free DIY guide" only matches that exact word order, so "DIY free fix" or "free diy plumbing" still slip through, and you end up maintaining a list that keeps growing. If DIY intent almost never converts for you, one single-word negative for DIY is simpler and covers more.

Checking for converting terms first is still the right move, but look at how many clicks sit behind them too. Two conversions from thirty clicks isn't enough to justify protecting "free" across the whole account.

Agreed, thirty clicks is too thin to judge anything. For local services there's another wrinkle: a lot of "free" queries end in a phone call, and if call tracking only counts calls over a minute, a quick "do you do free estimates?" call never registers as a conversion. So "free" can look dead when it's really just unmeasured. Before negating it, check whether those calls were ever being counted. Are calls set up as conversions at all?

Check the threshold setting first. Google Ads has a minimum call duration for calls from ads, which decides which calls count as conversions, and I believe the default is 60 seconds. A quick "do you do free estimates?" call falls under that, so it never shows up in the reports. You can lower the threshold in settings, but it usually only applies going forward, which means the historical search term data still reflects the old rule. So look at that setting before you judge "free" on its numbers.

If calls route through a third-party tracker, they usually get to Google Ads through a separate conversion import with its own duration rule, so short "free" calls can get dropped twice. Are these Google forwarding numbers or a third-party tracker?

It depends on which system actually records the call, so settle that before you touch any negatives. With Google forwarding numbers, the call counts as a conversion under the duration rule on that conversion action, and you can see which search term it came from. With a third-party tracker, the tracker may log the short call while Google only receives an imported conversion, and that import can drop calls the tracker counted. If short calls vanish on the tracker side, no Google setting will bring them back. The quickest check is to pull the tracker's call log for one week and compare it against Google's numbers for the same week.