I relaunched ads after a suspension in a niche where scammers undercut prices
The Sheffield branch of a London company: lockouts and lock changes. The old Google Ads account was suspended, there was no Maps listing, and competitors in search promised rock-bottom prices.
45%
of relevant enquiries became paid jobs in August
$94
cost per completed job in August, down from $151 in July
24
completed jobs in August
0
suspensions of the new account in three months
This case is anonymised: I do not disclose the company name. Reports and correspondence are available on request.
BusinessEmergency lockouts, lock changes, door repairs
AreaSheffield and nearby towns: Barnsley, Chesterfield, Doncaster, Worksop
ChannelGoogle Ads, search
My partJune — August 2026, via an agency
01 · Niche
An honest locksmith loses to whoever promises a lower price
Local search is full of scammers: they lure people with a low price and charge far more on site. Someone locked out has no time to work out who is genuine.
Challenge 1 · competitors
Undercutting
Clicks in the niche are expensive and getting pricier: $3–5 in June, $11–19 in July and August.
Challenge 2 · Google
Under suspicion
Because of the scammers, Google checks local services more strictly: business verification, risk of suspension.
Challenge 3 · trust
No Maps listing
The company is registered in London but works in Sheffield. The address verification code for Google Maps never arrived.
How competitors work: another firm's bill the client shared in the chat
£4,200
For replacing the mechanism in one door: call-out £120, “internal mechanism” £1,950, warranty £450, plus VAT.
02 · Relaunch
The old account was suspended, so we rebuilt everything from scratch
In spring the agency spent three weeks unable to launch: Google had suspended the client's previous account. I started over: a new laptop, a new login, a new account.
SpringThree weeks without a launch: account suspended
Late MayA new account from scratch
Early JuneBusiness verification with a UK phone number
Warm-up weekA $10 test campaign
Mid-JuneFull launch
A warm-up instead of a sprint
$10
A small campaign ran for a week. Google did not suspend it and the ads showed on target searches. Only then was the account handed to the agency and the budget switched on.
During the warm-up
Google Tag Manager on the site
conversions: call from the ad, click on the number, form
keywords, negative keywords, ads
03 · Launch
Three campaigns for three different customer situations
Someone locked out at night and someone who has decided to change a lock search differently and cost differently.
1
Emergency, Sheffield
Locked out right now. The main campaign, $30 a day.
2
Emergency, nearby towns
The same searches in Barnsley, Chesterfield, Doncaster and Worksop. $12.50 a day.
3
Lock changes
Not urgent: locks and door repairs. $7.50 a day.
Growth without sudden moves: budget +20% a week
Start, mid-June$50 a day
August$110 a day
A sharp budget jump disrupts bid strategy learning and draws review attention. Every change was gradual, with an eye on the risk of suspension.
Why I did not test two landing pages straight away
14 leads a week is too few for a test
A split would give 7 per variant: the test would drag on for months and the strategy would learn from crumbs.
Volume first, then A/B
I proposed raising the budget to 25–30+ leads a week and only then splitting traffic.
04 · Tracking
We counted completed jobs, not clicks or calls
A locksmith gets paid for an opened lock, not for a call. From July the weekly report got a “jobs completed” column based on the client's call log.
August: from enquiry to payment
57unique enquiries
53relevant: excluding car keys
24completed jobs
The account saw half the calls
6 of 13
A week in early August: 13 calls in the log, Google counted 6. Cost per lead in the account was $71.87, in reality $33.17. The rising “cost per lead” was a tracking error, not an ad failure.
What else the audit that week found
“24/7” did not match the ad schedule
The ads said “24 hour”, but they only ran during working hours. Evening and Sunday calls came in, just not from the ads.
The best day hit the budget cap
Monday gave the cheapest lead and used up the daily limit every time.
The biggest loss was not in the ads
Of 13 calls, 4 booked. Five fell through because the technician was busy. That is a scheduling issue, not a bidding one.
Reviews where people read them
Yell, not Trustpilot
Without a Maps listing, trust is lower. Trustpilot is expensive at this volume, and emergency customers rarely ask about reviews.
Target for September, meeting on 26 August
≈ $75
per completed job — but not at the cost of fewer orders.
05 · Result
Every second relevant enquiry became a paid job
Enquiry → job
45%
Average for August. The last week: 50%.
Cost per completed job
$151→$94
July → August. A relevant enquiry cost $42.
Suspensions
0
From the relaunch to the handover, while the budget doubled.
The client was happy and ready to scale. At the end of August I handed the project to agency colleagues with a working setup and a clear target for the next month.
My part
Relaunch
A new account from scratch, business verification, the warm-up and the launch of three campaigns.
Analytics
Tag Manager, conversions, a weekly report down to the completed job, an audit of call tracking.
Client
Weekly reports and meetings with the owner, decisions on budget, reviews and the next month's target.
Google AdsGoogle Tag Managerlocal servicesrelaunch after suspensionend-to-end funnel to paymentUK
What I would do differently
Check the account currency
The account was created in dollars, while a UK business counts in pounds. Google does not let you change the currency after creation.
Reconcile calls with the log from week one
The tracking gap was only found in August. Until then the strategy was learning from half the signal.
Align ads with the technician's schedule earlier
Instead of pausing ads on the technician's days off, an ad schedule agreed in advance.
The company name and people's names are not disclosed. Figures come from the weekly Google Ads reports and the client's call log for June — August 2026. Reports and correspondence are available on request.
