Exact Match Isn't Exact: 64% of Search Spend Went on Google's Own Matches (UK)
Of Search Spend on Google's Own Matches
Exact-match keywords sent 76% of their spend to searches that were not exact — at £81.96 a lead against £37.24
The Challenge
At campaign level this account looked well behaved: £126K of Search spend in the quarter, cost per enquiry near target, nothing visibly broken. The trouble with a report at that altitude is that averaging is the whole point of it. A set of 740 live keywords spread across regional campaigns can hide a single keyword matching hundreds of unrelated searches, the same keyword bidding against itself in six campaigns at once, and a match type that no longer means what its name says — while the blended figure stays perfectly acceptable. This audit was a deliberate drop in altitude: every keyword, every search Google actually charged for, and the gap between the two.
Our Approach
- Tagged every itemised search term by what Google matched it as — true exact, close variant of exact, close variant of phrase, broad, AI Max — rather than by the match type we had set, so the question became whether Google's interpretation of our exact match was working, not whether exact match was working.
- Compared cost per lead inside each of those buckets rather than across the account, because a blended cost per lead cannot tell you which half of the traffic is paying for the other half.
- Summed spend by keyword text across campaigns rather than within them. A keyword duplicated across regional campaigns only shows its real cost when the campaigns are added together, which no standard report does.
- Measured what the search terms report does not show, by setting keyword-level spend against itemised search-term spend — the difference is the part of the budget sitting behind searches Google withholds.
- Banded keywords by Quality Score inside the non-brand campaigns only, so brand keywords — which always score well and convert cheaply — could not flatter the comparison.
- Treated each finding as a claim to be checked against the account rather than a rule of thumb, including one we had previously argued the other way.
The Results
Window: 1 July to 22 September 2026, Search campaigns only, £126,163 of keyword-level spend.
Close variants took 64% of Search spend. Of the £84,666 Google itemised by search term, £53,844 went to searches it judged close enough to one of our keywords — close variants of exact (25%), close variants of phrase (30%) and AI Max (8%) — at £70.31 per lead. The £30,821 that matched keywords as written cost £49.25 per lead.
Inside exact-match keywords, 76% of the spend was not exact. Exact-match keywords spent £6,440 on the searches we actually chose, returning leads at £37.24, and £20,252 on Google's close variants of them, at £81.96 — 120% more per lead for traffic we did not pick.
One keyword matched 818 different searches. "domestic fencing" took £5,626 and returned leads at £110.46, roughly double the account average. £5.97 of that — one tenth of one per cent — went to the phrase itself. The rest went to "fence panels", "composite fencing", "garden fencing" and 815 others.
Half of all keyword spend sat on keywords duplicated across campaigns. 119 of the 740 live keywords ran in three or more regional campaigns simultaneously, accounting for £63,440 — 50% of keyword-level spend. No campaign report can surface this, because each campaign only ever reports its own share. Summed across its five campaigns, "metal fencing near me" cost £840 per lead.
A third of Search spend is never itemised. £84,666 of the £126,163 appears in the search terms report; the other £41,497 — 33% — sits behind searches Google does not disclose. An audit built only on the search terms report is reasoning about two thirds of the money.
AI Max cost £77.92 per lead against a Search average of £60.84, with 48% of its spend going to gate searches at £108.62 per lead rather than the fence searches the campaigns were built around.
A correction we owe the record. We have argued previously that Quality Score is a diagnostic rather than something to manage toward. On this account, inside the non-brand campaigns, that was wrong in a way that costs money: keywords scoring 1–4 returned leads at £71.78, those scoring 7–10 at £40.51 — a £31.27 gap per lead across £66,170 of spend. Quality Score was not the cause of the gap, but it was tracking something real, and we dismissed it too readily.
What we are not claiming. The tightening these findings led to was applied on 23–24 September 2026, too recent for a clean before-and-after read, and October is mixed with the account's normal seasonal decline. No saving is claimed here. What is published is the diagnosis and the figures it rests on; the measured outcome will be added when there is a comparable window to measure it against.
Want results like this for your business?
Get a free Google Ads audit and see exactly where your budget could be working harder.
Get a Free Account Audit