SEO experiments
Three English calculator sites as a laboratory of what search engines reward, what they punish, and why.
Context
Seventeen clicks a day from Bing, not one from Google. And it is the same site: datumcalc.com, calculators for CNC machining, which Google keeps parked around position 78. I swept search results in DataForSEO, hunting for fields where page one was held by weak aggregators. They looked winnable. I built three English calculator sites on them: kerfworks.org for wood and joinery, datumcalc.com for CNC machining, and brewersmath.com for homebrewing. Alongside them stood a Czech bet, sepsano.cz with legal templates. All Astro on Cloudflare, same author, same technology. None of them had any history to draw on. The experiment could begin.
What I did
Google toyed with them for a few weeks, letting them into the results and pulling them back out; datumcalc.com held position 10 to 15 for a while. Then the engine picked its winner. It let sepsano.cz grow and wrote the English trio off. There was no technical explanation. At first glance everything was in order, 83 % of datumcalc.com's pages got indexed and the others looked much the same. The catch was elsewhere - the top spots in these fields are held by the likes of metal suppliers and tooling makers, for whom a calculator brings in enquiries. Most of those businesses have built up very strong authority over the years - a quantity that is measured only with great difficulty. And my measurement had a flaw - I focused on the queries about the calculators themselves and underestimated the strength of old, fortified domains - which then lend that authority to their calculators. I did not manage to persuade Google to pick my new site instead. Finding out along the way that an experiment like this appears on the internet roughly once a week did not help either.
Where it stands
Where Google turned up its nose, Bing curiously took hold - search works a little differently there. All three sites have some partial wins on it and have, among other things, become data sources for AI Overviews. DatumCalc is the most successful of them - on weekdays, when machinists are in their shops, it pulls around 200 AI citations and an estimated 15 clicks. The other two not even a tenth of that. In the end I cannot call any of these sites a success, and it is likely that once their domain renewals run out I will let them drift into oblivion. Still, they are interesting to watch, and they have given me a valuable look into how the major search engines work.
Happy to share more detail on request.