The context
The school trains ice cream makers and pastry chefs in Milan. It has two types of courses. The first are for people who want to open an ice cream shop from scratch, and run from the business plan to production and management. The second are for professionals who want to specialise.
The site is in Italian and English, and enquiries also come from abroad, for example from Germany, Canada and Mexico. The school doesn't sell online: interested people fill in the form on the course page, and the school then calls them back. The main advertising channel is Google, with campaigns on the searches of people looking for a course.
Numbers that didn't add up
At the start of 2026 I took over the campaigns myself. The first month, February, was a shake-down: 15 enquiries, with €65.39 invested for each. That was too much for a course, even though one enrolment is worth far more.
There was also a second, less visible problem. Google counted as conversions things that weren't enquiries. Requests for directions and estimated store visits ended up in the same number as completed forms, and inflated it.
The Google campaigns, month by month
I worked on the search keywords, the ads and the course pages, which serve as landing pages. Every week I looked at the search terms, removed those that brought the merely curious and pushed those that brought enquiries.
| Month | Enquiries | Investment per enquiry |
|---|---|---|
| February 2026 | 15 | 65.39 € |
| March 2026 | 37 | 31.03 € |
| April 2026 | 27 | 25.65 € |
From there the investment per enquiry settled between €25 and €31 a month. The best week was 18 to 24 May, with 9 enquiries at €21.24 each.
Counting the real enquiries
In July I compared the conversions with the forms that actually came from the site. Out of 347 submissions, 68 were tests or internal contacts and 100 were spam. The real enquiries were 171. Since then the dashboard counts the real forms, with the date and the province of the sender.
We keep requests for directions separate. Google counted over 11,800 of them, and since July there has also been a campaign to bring people to the group's ice cream shops. They are a useful signal for the shops, but they aren't course enrolments, and that's why they don't go into the count of enquiries.
Read the month, not the week
With about two thousand impressions a week and few enquiries, the weekly number swings a lot: the same investment can bring one enquiry or twelve. That's why decisions are made on the month, which is a reliable figure. The week is there to show whether something has broken.
What this case teaches
1. A conversion is not an enquiry
The platforms also count directions, estimated visits and test forms. The number that matters is the real forms, checked one by one.
2. Investment per enquiry comes down by working on searches
Removing the searches of the merely curious and pushing those of people who want to enrol more than halved the investment in two months.
3. With small numbers, look at the month
One good or bad week tells you nothing. The month does.
Frequently asked questions about marketing for cooking and ice cream schools
How much do you invest for each enquiry about a course?
It depends on the course and the market. In this case, after the first shake-down month, the investment settled between €25 and €31 per enquiry, with a low of €25.65 in April 2026.
Why don't Google's conversions match the enquiries?
Because Google also counts actions that aren't enquiries, such as directions or estimated store visits. Here, out of 347 submitted forms, the real enquiries were 171: the rest were tests and spam.
Is it better to read results every week or every month?
With few contacts a week, the weekly figure swings too much. Decisions are made on the month, while the week is only there to tell you whether something has broken.


