Case study, a real account
A tuning garage in Lancashire: what the takeover audit of a running account found, and the first week on the honest count
A vehicle tuning and diesel particulate filter garage in Lancashire had run its own Google Ads for years, with five live campaigns and a steady phone. We were given access on 24 September 2026, read the account through the Ads API before touching anything, and made one change on 1 October: the count the bidding learns from. This page is the audit as it read, the change, and the days after it, with the verdict deliberately withheld until mid October. The garage is not named; the numbers are its own.
Written by Daniel Stoychev, who read the account and made the change. Figures from the Google Ads API and the garage's own Analytics, read on 24 September and 5 October 2026.
What was already right
Read from the API, not assumed
- All five live campaigns on presence targeting, Google Search only, no search partners and no Display, exact and phrase match only, automatic recommendations off, AI Max off.
- Every ad and sitelink address returning 200, with old addresses redirecting to the right pages.
- Campaign negatives that blocked none of the campaign's own keywords and none of the searches that convert; one campaign deliberately excluding the filter terms to keep that traffic in its own campaign.
- Quality scores on the spending keywords mostly five to eight, with expected click-through rate and ad relevance average or better.
That list matters because the usual takeover audit opens with a page of faults. This one did not. The owner had built a sound account, and the audit's job was to find what a sound account still loses.
What the audit found
Five findings, in the order they cost money
| Finding | Evidence | Effect |
|---|---|---|
| Every live campaign was budget-limited | Search impression share lost to budget between 37 and 46 percent; lost to rank between 15 and 37 | The cheapest clicks in the account were the ones the budget could not buy |
| An account-level list of 363 generic negatives, in broad match | The list contained book, repair, fix, check, where, home and free, among others | Buyers typing book a remap or dpf repair near me were blocked before the auction |
| Sitelinks pointing at the wrong pages | Price sitelinks to the contact page; service sitelinks to the about page | The click paid for a page that did not answer it |
| Price assets disagreeing with the garage's prices | Three services shown at figures the garage no longer charged | A promise in the ad the page then broke |
| Product searches leaking | A cleaning-product search took about £53 for 40 clicks and 2 conversions; retailer and brand product searches beside it | Shoppers for a bottle, not customers for a job |
Two of the first day's alarms were withdrawn the same evening after a deeper read: an old call asset and a stray sitelink sat on campaigns removed long ago and could not show. The audit says so, because a finding that is wrong costs the owner's trust more than it saves.
The one change, and why it came first
1 October 2026, about half past nine in the evening
Before any budget, keyword or ad was touched, the count the bidding learns from was changed. The account had been counting a mix of actions, some of them not enquiries. From 1 October the counted conversions are three tagged actions on the site, a phone tap, a form and a messenger tap, and calls from the ads lasting twenty seconds or more. Six negatives went on the same evening. Nothing else changed between 1 and 5 October; the change history confirms it. The reason it came first is that every other change would be judged by the count, and judging a budget rise by a count that includes non-enquiries is how accounts grow in the wrong direction.
The days after
Two windows, read with the verdict withheld
| Window | Spend | Clicks | Real contacts | Per day | Cost per contact |
|---|---|---|---|---|---|
| 24 to 30 September, old count still bidding | £372 | 353 | about 51 | 7.3 | about £7.30 |
| 2 to 4 October, new count bidding | £170 | 176 | about 18 | 6.0 | about £9.40 |
Three days, eighteen contacts and a weekend are too few to read a trend, and Google files late conversions under the click's date, so the second window will rise before it settles. What the days do show: no sign of damage, Analytics agreeing with the Ads count line by line, the calls from ads matching the twenty-second rule exactly, lost-to-budget falling to nothing and lost-to-rank rising as the bidding relearns on the honest count. Ninety-seven percent of the clicks are on phones. The verdict on cost per contact is for the middle of October, and this page will carry it then, whichever way it goes.
What was not done
And what waits
- No budget rise, although every campaign was budget-limited. A rise judged on the old count would have bought more of whatever the old count rewarded.
- No keyword or ad changes, and no touch to the sitelinks or the price assets until the garage's new site is live, because the pages they point at are changing.
- The negative list of 363 words is read with the owner before it is cut, because some of its words are his and some are a previous agency's.
- No revenue is claimed. The garage knows which calls became jobs; this page does not yet.
Where this leads
The pages that do this work
Say what the ads are for. The fee and the first three changes come back in writing.
Calls, sales, bookings or leads, roughly what you spend, and the part of Greater Manchester you sell to. Daniel Stoychev reads it himself and answers within one working day with the tier, the fee and what he would change first. Already spending? The free check reads your account in 72 hours.