8 Apr 2025
How it started
On 8 April 2025, Karel Jakoube asked Michal for help with Google Ads again.
Rather than guess, Michal started with a one-hour audit. On 11 April, Karel accepted the offer: ads management at $287 a month, plus a $478 analytics setup.
Until tracking went live on 16 April 2025, nobody could say where the inquiries were coming from.
Karel Jakoube, JATEP:Our original primitive website and the advertising we set up ourselves worked at least somehow.
Apr 2025
173 of 674 keywords were DIY or off-target
Before April 2025, the campaigns held 674 keywords, all of them broad match. Of those, 173 targeted people searching for a spray, a trap or how-to instructions, not a technician.
The ad group behind the ant ad (shown below) bid on “ants for sale”: a pest-control company paying to reach people who wanted to buy ants. What those keywords brought in is impossible to know, because there was no conversion tracking before April 2025.
May 2025
Click-through rate 1.5% → 10.4%
By May 2025 we had rebuilt the account: broad match out, phrase match in, one town and one service per keyword.
Click-through rate rose eight months before there was a new website: from 1.5% in April 2025 to 10.4% in May 2025, the first month after the rebuild, and 9.9% across August–December 2025, with the old site still in place.
To be clear: the rebuild changed who clicked, not what they found when they got to the site.
Rendered from the account export, translated from Czech.
Jun 2025 –
Jan 2026
“That does not add up for us at all”
Karel ordered a new website in June 2025 for $1,555. On 5 October, he wrote that he would finish the price list “within the week”. Neither side wrote again until Michal followed up on 9 November.
Meanwhile, the old site had a good December.
The account, December 2025
25 leads
at $21 per lead
Karel’s count, December 2025
3 jobs
$502 in revenue
$507 to Google and $287 to us: about $800 for three jobs worth $502.
Both are true. A lead is a call, a contact click or a form. A job is a paying customer, and Karel was the one paying. Calls are the closest thing to a job the account can see, and 73% of leads since July 2025 have been phone calls. The new website was meant to close that gap.
Jan 2026
“We need jobs”
Website #1 went live on the evening of 14 January. At 04:19 the next morning, it hit a WordPress fatal error. At 01:00 on 16 January, the SSL certificate failed. Michal was on it at 01:35, and the site was back up at 12:02.
January closed at $100 per lead, after $29 in November and $21 in December. It was the worst month since tracking began in April 2025. The setup couldn’t sustain a good month.
Feb 2026
“We have to stop this”
From 21 January, we cut bids hard and added negative keywords.
Michal didn’t fight it. From 17 February, he worked unpaid. On 26 February, Karel paused the whole account himself. He restarted it on 2 March, and on 4 March he wrote: “The graphs are nice but the conversions aren’t there.”
Mar – Apr
2026
“It is actually just an orange strip”
He was right, and we had built it. Website #1’s code still contained the WordPress theme’s demo placeholders: a US phone number, +1-800-456-478-23, and info@email.com.
Going back to the old site wasn’t an option. On 17 March, Karel looked at a simpler template and saw no difference.
The idea: rework his agency’s estate-agent template, one he knew sold. Go-ahead on 28 March, at Michal’s cost. Karel approved it on 7 April. Live on 10 April. Never billed.
Karel Jakoube, JATEP:I think it looks good - well, best on mobile, which is actually what we need.
May – Sep
2026
Monthly swings shrank from 4.9× to 1.4×
On 5 May, we cut the daily budget from $33 to $14 to test the new site. Cost per lead held, so we raised the budget on 4 June. Leads went 11, 18, 22, 23 from May to August, and spend doubled by July while cost per lead stayed flat. The results earned the bigger budget.
On 4 August, we trimmed bids by about 30% and cut the daily budget to $11. August’s $17 per lead came on $397 of spend, down from $521 in July.
Cost per lead by month, USD
Aug 2025 to Sep 2026. *September through the 22nd.
From August 2025 to March 2026, cost per lead ran from $21 to $100, and the eight monthly figures averaged $47. From May to August 2026, it ran from $17 to $24, and the four monthly figures averaged $22.
Michal’s view in September 2026 is that results now hold “regardless of the extreme seasonality of the business”. The caveat: that covers only May to 22 September 2026, four full months and one partial month, from late spring to the summer peak. No winter yet.
Sep 2026
Results so far
- $40 → $22Cost per lead, May–Aug 2026 vs the same months of 2025 (−45%)
- 11 → 23Leads a month, May to August 2026
- 4.9× → 1.4×Gap between the costliest and cheapest month: a budget you can plan
- 1.5% → 10.4%Click-through rate, April to May 2025, after the account rebuild
- $254 → $521Monthly spend, May to July 2026, while cost per lead stayed at $22–$24
- 674 → 285Keywords, all broad match before, phrase and exact now
Year-on-year figures: Google Ads API pull of 11 September 2026, 73 leads for May–Aug 2026 at $22 a lead. Monthly figures: campaign report of 22 September 2026, 74 leads for the same months (the one-conversion difference is reporting lag). Ratios and averages of the monthly figures use unrounded values.
Amounts are converted to US dollars at the 2026 average exchange rate. Website #2 went live at 01:01 on 10 April 2026 (Prague time). Figures from JATEP’s Google Ads account and the email record.
Questions
How long did it take?
Tracking in a week, the account rebuild in a month, and the website that worked a year later, after one that did not.
What did it cost?
$287 a month for ads management, $478 for the analytics setup, $1,555 for website #1. Website #2 was never billed.
Would this work for my business?
If your cost per lead jumps from month to month and your website is not off limits. Half of this fix was the page.
In closing
Two fixes, in order. The May 2025 rebuild fixed who clicked. The April 2026 website fixed what they found on their phones. Only then did cost per lead stop jumping, and it has held from May to 22 September so far. No winter yet.
Spend ran about $200 to $520 a month from September 2025 to August 2026. This is a small local business, and that is the point.
This isn’t for anyone who needs results within a month, or whose website is off limits. Half of this fix was the page.
In Karel’s words
Karel Jakoube, JATEP:Yes, it is better. Now and then someone calls, but I would say about once every 3 days on average.
Once every three days is about ten a month. The account counted 11 leads in May. For the first time, his count and ours agreed.
On 26 June he ordered a second website from Michal, for his new real-estate career.