How we make the pages

Where the numbers come from

Every number on a page has a source and a date. When there is nowhere to take a number from, the page says “no data” in its place. We never make a number up or fill in a gap by guessing.

• Demand is the number of searches a month for the niche’s queries, from search statistics, with the date it was measured. A page without measured demand is not published.
• Competitors are the sites in the top 10 search results for the niche’s queries: who is there, what they offer and at what price. Prices come from the competitors’ own pages, with the date we read them.
• Customer pain points come from reviews and discussions. Each quote is matched against its source word for word, and one that does not match is thrown out.
• Launch costs come from open material about the niche. A figure reaches the page only when 2 sources confirm it.
• Unit economics follow the category’s norms (what a customer costs to win, what they bring) until the niche has measurements of its own. Such numbers are marked as category norms, never as a market measurement.

What the AI idea check does

The check searches for data on the idea and assembles a report on demand, competitors, prices, customer pains, channels and the money. Its score out of 100 adds up the ratings on nine axes: the market, demand, pain, competition, uniqueness, complexity, pricing, unit economics and growth. Each report shows the weights in its “How the score is calculated” block.

• An axis with nothing to measure it gets “no data” and stays out of the score, its weight shared among the others. If too few axes are measured, there is no overall score at all.
• Every statement in the report points to a measured fact, and every quote is word for word.

What the published checks add up to, with the data file to download, is in our study of the idea checks.

The limits of the check

• A check is an estimate from open data on the day it was measured. It is not a revenue forecast and not a promise of success.
• Search demand shows interest, and interest is not sales: people who search will not all buy.
• The competitors in a report are the ones visible in search. Offline players and closed communities may not show up.
• Category norms are averages. Your city, your format and your price can differ.
• Models make mistakes. That is why each number carries a link: open the source before you put money in, and talk to your future customers.

Market and idea pages repeat this caveat. Their data is an estimate from open sources on the day of measurement, and it is neither a promise of income nor financial or legal advice. Check the numbers for your own city and talk to customers before you invest. The decision is yours.

What AI writes and what a person reviews

• Guides, the glossary and the templates. Our agent Fox drafts them from data and real checks. The editor personally reviews every guide right after it is published: he checks the numbers against the sources, edits the text and signs the guide, so every guide carries the editor’s name and the date of the review.
• Market and idea pages are assembled by AI from measured data and real checks. Before publication each one passes automatic checks: every number has a source, demand is measured, there is enough data and the page does not repeat another one. An editor reads a sample from every batch, and a page that fails the checks is not published.
• Photos of people on the site are real, never AI-generated. An illustration drawn by AI is captioned as one. The audio version of a guide is speech synthesis, and the player says so.

How we score company websites

Our niche pages list competitors’ websites, and some of them have a website audit page. These are measurements, not an opinion about the company.

• We score the first three direct competitors on the niche page, in the order they rank in search for the niche’s queries. Sites on website builders, marketplaces, social networks and directories are not scored.
• We audit a site only where its robots.txt allows it: if the file disallows crawlers, we leave the site alone.
• We measure up to 25 pages of the site: indexing, speed, titles and markup, content, the path to an enquiry, trust (a secure connection, contacts, a privacy policy) and search visibility for the niche’s queries. Each area scores 0 to 10; the total is their weighted sum out of 100.
• The offer stays out of the score: a model reads it, which is not a measurement. Registry data about a company (court cases, debts, finances) is never shown on a public page.
• An area with nothing to measure it (the site did not answer, no data) is “not measured” and stays out of the score. With less than 60% of the weight measured there is no total.
• Every audit page shows the day it was measured: it is the site as it was that day. The site’s owner can run the audit again for free at any time.
• We publish only the pages scoring 60 out of 100 or more. Other results stay internal and are not on the site.
• If it is your site, you can run the audit again for free on its audit page and write to us: we correct a wrong measurement or remove the page at the site owner’s request.

How often the data is refreshed

• Demand: once a month.
• Search results, launch costs and prices: once a quarter.
• The idea checks on the pages: every 6 months, or sooner if demand moves by more than 30 %.
• A refresh erases nothing. When a new measurement fails, the previous data stays with its own date.
• A page nobody has found in search for 90 days gets reviewed: we merge it with a close one or take it down.

About the use of AI

We label AI-made work ourselves. Each page that AI assembled or wrote says how it was made and who reviewed it. If you see a mistake, press “Report an error” at the bottom of the page; how we fix mistakes is set out in the editorial policy.