Most sites publish their wins. This page publishes the conditions under which this project stops.
Calceva was launched as a stated trial, with its bar set before measurement began. The threshold below is not a goal we hope to reach; it is the line under which the project shuts down and the pages come down rather than rot. We publish it because a threshold announced afterwards is a threshold that gets negotiated.
The bar, and the clocks
2027-02-06
A second floor, added 20 August 2026 — and the first one stands. We are not moving the 437-impression bar; we are adding a harder one beside it. By the same date and over the same rolling 28 days, Search Console must also show a median of at least 35 clicks a day. If either floor is missed, the project stops. The figure is not a new guess: it is the same 437 impressions at the 8% click-through rate our own revenue model assumes. We added it because impressions can rise while clicks do not: in the 28 days to 17 August 2026, our first weeks online, Search Console counted 1,699 impressions and 4 clicks. A threshold that a site can clear without a single visitor is not a threshold.
What this threshold does not measure: whether a generative engine quotes us. Search Console reports Google, including its AI Overviews, but not ChatGPT, Perplexity or Claude — and no honest counter exists for those, because a large share of their referrals arrive carrying no referrer at all. So we judge that work by citation, tested directly, and we leave this number where we set it rather than move the goalposts to a measure that would flatter us.
And the citation work is itself on two clocks, stated before they start. As of 9 August 2026, Bing, DuckDuckGo and Brave — the indexes generative assistants actually browse — carry zero pages of this site. So: (1) within 30 days of the first crawl attempt by bingbot visible in Bing Webmaster Tools after the sitemap is submitted, Bing must index something, or the machine-readable effort is declared dead at the root; (2) within 60 days of the first day Bing Webmaster Tools reports at least one calceva.com page in Bing's index, our fixed panel must produce at least one answer that names or links this site, or the effort is dropped as an acquisition channel. That count runs on the 12 long-tail questions only — the ones where our pages are plausibly the source. The 4 brand questions and the 4 head-term questions stay in the record as diagnostics and are not counted: asking an assistant about “calceva.com” and being told about calceva.com measures nothing, and the head terms are the negative control. We narrowed this on 25 August, after finding that every citation we had recorded came from the brand questions — a threshold that cannot fail is not a threshold. The indexation clock started on 2026-08-09: Bing Webmaster Tools is connected, the sitemap submitted, and bingbot’s crawl attempts are visible (last: 5 August). The citation clock started on 2026-08-25: on that day Bing Webmaster Tools reported 3 pages of this site in Bing’s index, which is the trigger we set for it. Two things about that date, because the page should not claim more precision than it has. It is the first reading we confirmed, not necessarily the first day it was true: the real date lies between 9 and 25 August, so this deadline is generous to us by up to sixteen days — and for a clock whose failure ends an effort, erring towards more time is the cautious direction, not the flattering one. And a site: query on Bing still returns nothing for us, which is why we read the index itself rather than the results page: they disagree, and only one of them is the index. A deadline without a start condition is a wish, and a start condition without a deadline is a stall.
What this page is not
None of this changes what the calculators compute or how the numbers are verified. The accuracy machinery — sources, dates, tests, replays — is documented on the methodology and reliability pages, and it holds regardless of how this trial ends.