← marketing AndreawithAI 02 / 05 IT

Technical training: from wrong clicks to booked appointments.

A technical training centre was paying more than a hundred euros for every lead. In three weeks, with daily spend unchanged, the cost fell to a sixth: wrong keywords stopped, offer changed, one landing page for each course.

sector
vocational technical training: welding, CNC machines, CAD drawing
channels
Google Ads (search) and Meta Ads
goal
forms completed and appointments booked from the course pages
constraint
no promises of earnings in the ads or on the pages

01Where it started

Search campaigns on Google, one for each family of courses, with the "maximise conversions" bid strategy and no limit at all. With one conversion a week the algorithm has no data to work with: after five leads in four days it raised its bids to €3.62 per click and spent on informational searches, such as people looking for a licence or a manual. The result: a week at zero, without anyone having touched the campaign. On Meta the forms were coming in, but all the ads pointed to the same generic page.

02What I did

  1. I read the search terms, one by one. Half the clicks in one campaign came from people who already use programming software and were looking for simulators, manuals, code: not a course. Phrase-match negative keywords, every week.
  2. I changed the bid strategy. From "maximise conversions" to "maximise clicks" with a maximum cost per click: with little data, an explicit limit is worth more than the algorithm. The cost per click fell from €3.31 to €1.16.
  3. One landing page for each course. A dedicated page, form visible straight away, the course's own text and no promise of income. On Meta the structure was rebuilt into two hubs, one ad per course on its own page.
  4. I reconciled the conversions with the client's CRM. Google counted 5 appointments in a week, the CRM one: the count was switched to "one conversion per click" and every week the forms are checked against the client's spreadsheet, to the euro.
  5. Three changes in four days, read separately. Location, budget and landing page changed in sequence; in the days after I looked at what had moved and when: the cost per click had already dropped with the first change, the form completion rate only with the new page.

03What worked, and what didn't

  • The dedicated landing page: clicks that turn into forms went from about 4% to about 9%, and the cost per form from €83.82 to €16.81.
  • The cap on cost per click: healthier auctions, same spend, four times the conversions.
  • On Meta, the page for each course in place of the generic one: from €45.87 to €11.27 per lead in a week.
  • The automatic bid strategy with no limits: a whole week with zero conversions, caused by the algorithm and not by any change.
  • A single keyword spent 39% of its campaign in eight days without a conversion, before I stopped it.
  • For a week the clicks on one campaign stopped converting with no fault anywhere: on mobile the page opened two cookie banners, one on top of the other. The client fixes that, not the campaign.

04Results and source

The account's cost per conversion went from €102.71 to €31.24 to €17.43 over three consecutive weeks, with the same daily spend. In the last week: 20 forms completed and 13 appointments booked. The figures come from the Google Ads API in read-only mode, conversions attributed to the click date, read on 20 October 2026; the Meta ones from Ads Manager, seven days from 23 to 29 September.

Cost per conversion, week by week
Google Ads · conversions by click date · daily spend unchanged
102.7128 Sep → 4 Oct
31.245 → 11 Oct
17.4312 → 18 Oct
€ per conversion102.71 → 17.43, −83%
−83%cost per conversion in three weeks€102.71 → €17.43, Google Ads, weeks 28 Sep → 18 Oct
2×clicks that turn into formsfrom 4% to about 9%, on the dedicated landing page
−75%cost per lead on Meta€45.87 → €11.27, generic page against course page, 7 days
13appointments booked in a week12 → 18 Oct, count reconciled with the client's CRM

The missing number. How many of these leads become paying students only the client knows, and it is not yet in my data. Until it is, the cost per conversion is a clue, not the result.

updated on 26 Oct 2026