I called a keyword set off-product before I opened the client's shop: how to check product–query fit before cutting a Google Ads campaign
I wrote “wrong market fit” into a weekly ads review about seven campaigns burning money at 0.22 ROAS. Then I opened the client's shop and found five products in exactly that category. The four-minute check that would have caught it.
I wrote this sentence into a client's weekly ads review on Monday: "That intent gap — not bid management — is why €5,314 produced two conversions." It was the strategic conclusion of the whole document. Seven new international Search campaigns had burned nearly five thousand euros in a week and returned a ROAS of 0.22, and I had just explained why: people were searching for complete electric bed bases, and the shop sells a retrofit lift device. Wrong market. Consider killing the campaigns.
Then I opened the client's own shop.
They sell five products in that category. Two-zone electric slatted base. Seven-zone electric slatted base. One with a standing aid. An electric adjustable mattress. And the retrofit lift I had built my entire argument around. The homepage title tag says, in plain German, that this is a shop for electric slatted bases and mattresses. The French storefront carries the same full range and has a decision tree on the page that routes "my bed base is at the end of its life" straight to the mid-priced model.
So elektrischer Lattenrost and sommier électrique 140x190 were not mismatched queries. They were on-product queries. I had diagnosed a product-market problem in a business that sells the product.
What I got right, and what that hid
The awkward part is that the recommendation was still correct. A 0.22 ROAS is a 0.22 ROAS. Keyword expansion was matching the ads to competitor brand names and to mattresses and pillows the client does not sell in those ad groups. One click on a payment-intent query cost €59.76. Pausing that build-out was the right call on Monday and it is still the right call now.
Being right about the action is exactly what let me be wrong about the cause for a week. Nobody pushes back on a recommendation they agree with. The budget number moves, everyone nods, and the reasoning underneath it goes unread — including by me.
And the wrong cause leads somewhere different. "Wrong market fit, consider killing" points at abandoning four countries. "Right demand, wrong routing" points at rebuilding with per-product ad groups and landing pages that match the query. Those are opposite plans with the same first step, which is the most dangerous kind of mistake in this job, because the first step goes fine.
It also explains something I had filed as a separate oddity. The German text campaign was running an ad that pitched the €219 retrofit instead of a slatted base — to people searching for a slatted base — while the same company had three of them on the shelf at higher prices. That ad was not sloppy copywriting. It was the same premise error, written by someone else, months earlier, and I had walked straight past it.
In Ian Leslie's How to Disagree, Socrates tells Callicles that if he refutes him he will "go on record as my greatest benefactor." That is the correct attitude to a correction, and it is much easier to hold when the person doing the refuting is you, on a Tuesday, with nobody watching.
The check that would have caught it, in about four minutes
Before you call a keyword set off-product, open the shop and list what is actually for sale in that category, at what price, on which URL. Then take the top ten search terms by cost and write the matching product next to each one. Terms with no product are the real mismatch. Terms with a product that the ad never mentions are a routing problem, and routing problems are fixable at a fraction of the cost of abandoning a market.
That second bucket is where this account was living. Ads and landing pages were pointing at the cheapest item in the range while the searcher had described a different, more expensive one. Google's own Quality Score documentation puts landing page experience and ad relevance alongside expected click-through rate for exactly this reason — the auction is already telling you when the destination does not answer the query. I read that signal as evidence of a category mismatch when it was evidence of a mapping mistake.
The tooling makes this easier to get wrong than it used to be. AI Max for Search expands beyond your keywords with what Google calls keyword-less matching, and pairs it with final URL expansion, which picks a landing page for you. Google's own announcement leads with the upside, and the upside is real when the underlying product mapping is sound. When it is not, the system industriously buys traffic for products you do not stock and sends the rest to the wrong page, and the search-term report that lands on your desk looks exactly like a market that does not want you.
Why this is the same error I keep writing about
I have made a version of this mistake in public before. I diagnosed nine months of missing purchase events without access to the tag manager and had to verify an inherited premise the hard way. I watched a channel filter quietly break a ROAS calculation because the numerator and denominator were counted from different places. Every time, the data was fine and the frame around it was not.
The pattern is that I trust a structured export over an unstructured look. The search-term report is a table; the shop is a website. One arrives ranked by cost and the other requires me to click around like a customer. So I read the table, and the table has no column for "does this company sell this."
The counterweight is the thing I keep coming back to as the actual leverage in this work: domain knowledge, not tooling. Except domain knowledge is not a feeling of familiarity. It is having looked, recently, at what the client sells. I had been on this account for weeks and I had absorbed the flagship product as though it were the whole catalogue, and then reasoned confidently from a range of one.
I filed a correction note to the client the same day, before anyone asked. The rebuild is now three product-matched ad groups, each pointing at its own product page, expansion off, phrase and exact only. It might have been three markets written off instead.
Read the shop before you read the report. The report will never tell you what is missing from it.
Sources & further reading
External
How AI Max for Search campaigns works — Google Ads Help
About Final URL expansion in Search — Google Ads Help
Unlock next-level performance with AI Max for Search campaigns — Google
About Quality Score for Search campaigns — Google Ads Help
Related posts
No login to the tag manager
The channel filter on my dashboard silently broke ROAS
How to read a competitor's channel mix from the outside
The death of generic AI