We spent DKK 709 on Google Ads. Nobody signed up.
In two and a half days, 759 people saw our ad, 20 clicked, and none of them made an account. Here is every figure, from the ad account and from our own analytics, what went wrong, and what we would do first next time.
- 759 people saw the ad, 20 clicked, it cost DKK 709.10 (about €95), and nobody signed up.
- A click cost DKK 35, more than twice the DKK 15 we planned, and almost half the money went on one phrase that matched other companies' products.
- In the same days our YouTube Shorts sent 49 visitors for nothing. Ads are now last on our list, not first.
On 21 October we turned on Google Ads for tidekeeper, the web analytics we make in Denmark. Two search campaigns: one for people in Europe looking for analytics that respects their visitors, one for developers and their AI agents, in Europe and four English-speaking countries. On the evening of 23 October we paused both. This is the receipt.1
Where the 20 clicks went
We don't put a Google tag on our pages: they count visitors without a cookie, and a tag would undo that. So we read the clicks the way our customers would, with the campaign written into each link and counted in tidekeeper.2
Nineteen of twenty arrived, which is about as good as counting gets. Then they left. Most landed on the page about our MCP server, the rest on the front page and pricing.
| Entry page | Visits | Left after one page | Time on site |
|---|---|---|---|
/docs/mcp | 10 | 70% | 12 s |
/ | 4 | 100% | 0 s* |
/pricing | 3 | 100% | 0 s* |
* A visit of one page has no second view to time it by, so it shows 0 s: a quick look and a long read look the same. That one is ours to fix.
Almost half the money went on one phrase
Our best guess was that people building with AI agents would search for analytics they can plug in. The phrase mcp server analytics got 10 of the 20 clicks, and DKK 335.24 of the money.
Then we read what people had actually typed. Google stretches a phrase to its "close variants", so the ad showed for other companies' MCP servers too. People looking for someone else's server spent 12 seconds on ours, on average.
A click cost DKK 35, not the 15 we planned
We planned to pay at most DKK 15 a click. Two settings changed after launch: the limit came off one campaign, and the other was switched to "Maximize conversions". Without a Google tag, Google sees no conversions, so it was maximizing nothing, and it marked the campaign "Limited" for exactly that reason.
The rest was what the searches told us. Bare competitor names, plausible or matomo analytics, are people on their way to those sites, and each of those clicks cost DKK 22 to 46. Nobody searches for analytics in a terminal: that ad group was never shown once. And the lesson for us: read the live campaign, not the plan you wrote. Google's set-up screens fill in their own suggestions, and the summary didn't always match the steps.
The same days, free did better
At DKK 35 a click, and a few in a hundred signing up, one customer would have cost well over DKK 1,000. Our plans start at €9 a month. None of this is new. Dropbox found the same in 2009: search ads cost them $233 to $388 for each paying customer, for a product that cost $99 a year.3
SEM is a way to harvest demand, not create it.
What we would do first, next time
Ads aren't the villain. They are an amplifier: they make louder what already works. We turned one up on nothing. In order, then:
- Count first. Know where every visitor comes from, and which ones become customers. This is the part we make.
- Make one page worth finding. The best answer to a question your customers already ask, for Google and for the AI that answers them. This post is ours.
- Build a loop. Give people a reason to bring the next person.
- Show up as a person. Post, film, answer. It adds up, and nobody bills you per view.
- Then pay. For what already works, with a limit on every click.
- Our Google Ads account, both campaigns, read on 23 October 2026 when we paused them.
- tidekeeper.eu's own figures, 21 to 23 October 2026: sources, entry pages and the signed-up event.
- Drew Houston, Dropbox: Startup Lessons Learned (2010), slides on SlideShare.
- Eugene Wei, Learnings from Dropbox (2010), quoting Houston's talk.
# We spent DKK 709 on Google Ads. Nobody signed up. > tidekeeper's blog, 27 October 2026, by Christina Sørensen: https://tidekeeper.eu/blog/google-ads-receipt In two and a half days, 759 people saw our ad, 20 clicked, and none of them made an account. Here is every figure, from the ad account and from our own analytics, what went wrong, and what we would do first next time. ## In short - 759 people saw the ad, 20 clicked, it cost DKK 709.10 (about €95), and nobody signed up. - A click cost DKK 35, more than twice the DKK 15 we planned, and almost half the money went on one phrase that matched other companies' products. - In the same days our YouTube Shorts sent 49 visitors for nothing. Ads are now last on our list, not first. On 21 October we turned on Google Ads for tidekeeper, the web analytics we make in Denmark. Two search campaigns: one for people in Europe looking for analytics that respects their visitors, one for developers and their AI agents, in Europe and four English-speaking countries. On the evening of 23 October we paused both. This is the receipt.[1] | | | |---|---| | Saw the ad | 759 (impressions) | | Clicked | 20 (2.6% of them) | | Cost | DKK 709.10 (about €95) | | Signed up | 0 (counted on our side) | | Average cost of a click | DKK 35.46 | | Developers and agents campaign | DKK 429.77 | | The limit we had planned | DKK 15.00 | | Privacy campaign, Europe | DKK 279.33 | Figure 1. Google Ads, 21 to 23 October 2026, both campaigns. Source: our Google Ads account, read when we paused the campaigns; sign-ups counted in tidekeeper. ## Where the 20 clicks went We don't put a Google tag on our pages: they count visitors without a cookie, and a tag would undo that. So we read the clicks the way our customers would, with the campaign written into each link and counted in tidekeeper.[2] 1. 759: saw the ad 2. 20: clicked it 3. 19: arrived on tidekeeper.eu: 17 with our tags, 2 without 4. 0: made an account Figure 2. From the ad to an account. Source: Google Ads for the first two; tidekeeper.eu's own figures for the rest. Nineteen of twenty arrived, which is about as good as counting gets. Then they left. Most landed on the page about our MCP server, the rest on the front page and pricing. | Entry page | Visits | Left after one page | Time on site | |---|---|---|---| | `/docs/mcp` | 10 | 70% | 12 s | | `/` | 4 | 100% | 0 s* | | `/pricing` | 3 | 100% | 0 s* | * A visit of one page has no second view to time it by, so it shows 0 s: a quick look and a long read look the same. That one is ours to fix. Figure 3. Where the 17 tagged visits landed. Source: tidekeeper.eu's own figures, 21 to 23 October 2026. ## Almost half the money went on one phrase Our best guess was that people building with AI agents would search for analytics they can plug in. The phrase `mcp server analytics` got 10 of the 20 clicks, and DKK 335.24 of the money. | | | | |---|---|---| | mcp server analytics | DKK 335.24 | 10 clicks | | simple web analytics | DKK 142.05 | 4 clicks | | analytics mcp | DKK 94.53 | 2 clicks | | plausible analytics alternative | DKK 72.17 | 2 clicks | | google analytics alternative | DKK 37.75 | 1 click | | cookieless analytics | DKK 27.36 | 1 click | 47% of the money, and half the clicks, on one phrase. Figure 4. Where the DKK 709.10 went, by keyword. Source: our Google Ads account, phrase match, both campaigns. Then we read what people had actually typed. Google stretches a phrase to its "close variants", so the ad showed for other companies' MCP servers too. People looking for someone else's server spent 12 seconds on ours, on average. - website analytics (×2) - databricks mcp server (×2), looking for someone else - plausible analytics alternative (×1) - amplitude mcp server (×1), looking for someone else - plausible (×1), looking for someone else - plausible analytics (×1), looking for someone else - matomo analytics (×1), looking for someone else - mcp gsc by … (×1), looking for someone else Figure 5. What people typed before they clicked; ringed in orange, someone looking for something else. Source: our Google Ads account's search terms, which show the search for about half the clicks. ## A click cost DKK 35, not the 15 we planned We planned to pay at most DKK 15 a click. Two settings changed after launch: the limit came off one campaign, and the other was switched to "Maximize conversions". Without a Google tag, Google sees no conversions, so it was maximizing nothing, and it marked the campaign "Limited" for exactly that reason. The rest was what the searches told us. Bare competitor names, `plausible` or `matomo analytics`, are people on their way to those sites, and each of those clicks cost DKK 22 to 46. Nobody searches for analytics in a terminal: that ad group was never shown once. And the lesson for us: read the live campaign, not the plan you wrote. Google's set-up screens fill in their own suggestions, and the summary didn't always match the steps. ## The same days, free did better | | | | | |---|---|---|---| | Google Ads | 19 | for DKK 709.10 | 0 sign-ups. | | Our YouTube Shorts | 49 | for DKK 0 | The one sign-up that week came from the US with no source, so not from an ad. | Figure 6. Visitors who arrived, 21 to 23 October: one dot, one visitor. Source: tidekeeper.eu's own figures, 21 to 23 October 2026. At DKK 35 a click, and a few in a hundred signing up, one customer would have cost well over DKK 1,000. Our plans start at €9 a month. None of this is new. Dropbox found the same in 2009: search ads cost them $233 to $388 for each paying customer, for a product that cost $99 a year.[3] > SEM is a way to harvest demand, not create it. > > — Drew Houston, Dropbox, Startup Lessons Learned, 2010[4] ## What we would do first, next time Ads aren't the villain. They are an amplifier: they make louder what already works. We turned one up on nothing. In order, then: 1. **Count first.** Know where every visitor comes from, and which ones become customers. This is the part we make. 2. **Make one page worth finding.** The best answer to a question your customers already ask, for Google and for the AI that answers them. This post is ours. 3. **Build a loop.** Give people a reason to bring the next person. 4. **Show up as a person.** Post, film, answer. It adds up, and nobody bills you per view. 5. **Then pay.** For what already works, with a limit on every click. The film, Don't Use Google Ads. Do This Instead. (8:02): https://youtu.be/rkFOHK9cZAk ## Sources 1. Our Google Ads account, both campaigns, read on 23 October 2026 when we paused them. 2. tidekeeper.eu's own figures, 21 to 23 October 2026: sources, entry pages and the signed-up event. 3. Drew Houston, [Dropbox: Startup Lessons Learned](https://www.slideshare.net/slideshow/dropbox-startup-lessons-learned-3836587/3836587) (2010), slides on SlideShare. 4. Eugene Wei, [Learnings from Dropbox](https://www.eugenewei.com/imported-mt-blog/2010/4/28/learnings-from-dropbox.html) (2010), quoting Houston's talk.