A commercial HVAC company in Sacramento handed me an account spending roughly $14,000 a month. Before touching anything I did what I do on every Google Ads audit: exported the search terms report for the previous 90 days and sorted it by cost.
Third on the list, by spend, was a query about HVAC technician salaries. Also in the top twenty: several queries about how to fix a residential air conditioner yourself, two about HVAC certification courses, and one that was clearly a student researching a coursework assignment. Together these accounted for a bit over a fifth of the spend in the period. Nobody had opened that report in seven months.
The agency before us had already drawn their conclusion and acted on it: they had moved everything to phrase match to "stop the waste". Volume collapsed, cost per lead went up, and the waste did not actually stop, because the problem was never the match type.
What broad match is actually reading
People talk about broad match as if it still works the way it did in 2015, matching on loose synonyms and ruining accounts. It does not. Broad match today is interpreted by the bidding system using several inputs at once: the keyword itself, the other keywords in the ad group, your landing page content, your account's conversion history, and any audience signals you have supplied.
That means broad match is only as good as the signals feeding it. Give it a clean conversion action tied to real revenue, a relevant landing page and a sensible negative list, and it will find queries your keyword research never would. Give it a conversion action that fires on every form submission, a generic homepage, and no negatives, and it will enthusiastically buy you salary searches.
Switching to phrase match without fixing the signals just makes a badly-informed system reach fewer people. You get a smaller version of the same problem, and you lose the queries that were genuinely profitable.
The audit, step by step
Export the search terms report for the last 90 days with cost, conversions and conversion value. Sort by cost descending. Then tag every term that spent anything meaningful into one of five buckets.
| Bucket | What it looks like | Action |
|---|---|---|
| On target | Commercial intent for what you sell, in your area | Keep, consider adding as exact |
| Adjacent | A service you do not offer, or a segment you do not serve | Negative, or build a page if it is worth serving |
| Research | How to, DIY, what is, tutorial, guide | Negative at account level |
| Employment and education | Jobs, salary, hiring, certification, course, training | Negative at account level |
| Wrong entirely | Other industries, other countries, misreadings | Negative, and check your geo settings |
Add up the spend in the bottom three buckets. In most accounts I audit this sits somewhere between 20 and 45 percent of total spend. That figure is your immediate, recoverable budget, and it is usually larger than anything you would gain from a month of bid tinkering.
Building negative lists that keep working
One-off negatives added in a panic do not scale. Build shared lists at account level and apply them to every search campaign, so new campaigns inherit the protection automatically.
I keep four standing lists on most accounts:
Employment and education. Jobs, careers, salary, hiring, apprentice, internship, course, certification, training, degree, exam. This one list alone usually recovers several percent of spend for any skilled-trade, legal, medical or technical advertiser.
DIY and research. How to, DIY, yourself, tutorial, guide, meaning, definition, what is, examples, template, free. Be careful with "free" if you genuinely offer a free consultation; in that case keep it out of the list and handle it at ad group level.
Price-shopper terms that do not fit your model. Cheap, cheapest, budget, discount, second hand, used. Only if you are a premium provider. If you compete on price, these are your best queries and you should be bidding on them deliberately.
Irrelevant verticals. Build this one from your own report over time. Every account accumulates its own strange set.
Then one more, specific to your business: competitor brand terms, but only after you have decided deliberately whether to bid on them. They are expensive, they convert at a lower rate, and for some businesses they are still worth it. The mistake is letting broad match buy them accidentally and never noticing.
Why the conversion action matters more than any of this
If your conversion action is "any form submitted", the system will optimise toward the cheapest possible form submission, which means it will keep finding low-intent queries even after you have negated the obvious ones. You will be playing whack-a-mole forever.
Fix the signal first. Use one primary conversion action that corresponds to a qualified lead or a booked appointment, import offline conversions from your CRM if you can, and let the bidding learn from outcomes that are worth money. Once the system knows what a good customer looks like, broad match becomes an asset rather than a liability, because it starts hunting for people like the ones who actually paid you.
The cadence
Weekly for the first month, then every two weeks once the obvious waste is gone. It takes about twenty minutes once the lists exist. Put it in a calendar, because this is the single highest-return recurring task in a search account and it is also the easiest one to let slide for seven months.
For the Sacramento account the first pass negated about 19 percent of spend, which went back into the campaigns that were producing booked site visits. We kept broad match. Within six weeks the cost per booked job was down meaningfully and the volume was higher than it had been on phrase match, because the system could finally explore without wandering into the careers section of the internet.
The conversion-action point deserves its own read: why cheap Google Ads leads are often the useless ones. If you also run Performance Max, check that it is not quietly doing the same thing with your brand terms, covered in Performance Max cannibalising brand search. And if your reported conversions look nothing like your actual orders, that is a separate and very common problem.
If you want someone to read your search terms properly and tell you what share of your spend is going nowhere, the free growth audit covers exactly that, and you get the list of wasted queries in writing. Our Google Ads work runs this every week.