How Kateion finds and checks buying signals
Kateion collects public posts from X, the open web and public forums, uses a language model plus fixed rules to decide whether each one is a real business at a buying moment, scores it for fit, timing, intent, confidence and freshness, and has a person check it before it goes into a buyer’s weekly list.
1. Listen: public sources only
Each campaign runs missions aimed at one kind of moment for one kind of business. Missions search:
- X, through xAI’s search. Only posts that appear in the search citations are kept, so every lead has a real, linkable source.
- The open web, through search engines and our own crawler. The crawler respects robots.txt and never visits private hosts.
- Public forums and community threads that show up in ordinary web search.
- Watched websites, re-checked weekly so that a change, such as a vendor swapped out or a shutdown notice, becomes a signal.
2. Read: is this a business at a moment?
Every new item is read once by a language model and then finalised by fixed rules. The questions are always the same:
| Question | Why it matters |
|---|---|
| Is it a real business or a consumer? | Consumers are never routed to a buyer. |
| What changed, and with which vendor? | Price change, outage, deprecation, licence change, acquisition, policy change, support collapse, or openly shopping for an alternative. |
| Who is it? | Only from what the business made public: a website in the bio, a named company. Anonymous accounts are never de-anonymised. |
| Which buyer is it for? | Routed by campaign rules, not guesswork. |
3. Score: the same way every time
Each candidate gets one score out of 100 built from five parts:
- Fit: is this the kind of business the buyer serves?
- Trigger: how strong is the moment (a price that doubled beats a mild complaint)?
- Intent: are they actually looking to change?
- Confidence: how sure are we about the reading and the identity?
- Freshness: how recent is it? Most moments go cold within two or three weeks.
Minimum thresholds on fit, trigger or intent, confidence and identity keep weak items out entirely, however high the rest scores.
4. Check: a person decides
A person reviews every candidate and marks it good, bad or worth watching. Only items marked good go into a buyer’s list. Buyers tell us which leads were useful, contacted or won, and that feedback tunes the scoring.
5. Deliver: one list a week
Each buyer gets one short list a week on the same day: company, moment, why now, how we know who it is, and the link to the post. Every delivery is recorded, so the same lead is not sent twice and a niche can be offered exclusively.
The rules
- Public posts only. Only what a business wrote about itself, in public. No private data, no bought lists.
- Link, don’t copy. Every entry links to the original post. We quote a line at most.
- A person at every contact. Nothing is sent automatically, by us or for you. People write to people.
- Removed on request, same day. Anyone can ask to be taken out, and every buyer is asked to honour it.
We keep the minimum: the link, a short quote and the scores. Source items are deleted after 180 days. If your post appears in a list and you want it out, tell us and it is removed the same day.
About the method
How does Kateion find buying signals?
Kateion collects public posts from X, the open web and public forums, uses a language model plus fixed rules to decide whether each one is a real business at a buying moment, scores it for fit, timing, intent, confidence and freshness, and has a person check it before it goes into a buyer’s weekly list.
Which sources does Kateion use?
Public posts on X (through xAI’s search, keeping only posts it can cite), the open web through search engines and our own crawler that respects robots.txt, and public forums and community threads. Reddit’s API is not used.
Does a person check every lead?
Yes. Software finds and scores candidates; a person reviews each one before it is delivered, and nothing is ever sent to the business automatically.
How long is data kept?
Only what is needed: the link, a short quote and the scores. Source items are deleted after 180 days, and anyone can ask to be removed sooner.
How is each lead scored?
On five things, the same way every time: fit (is this a business you sell to), timing (how strong the moment is), intent (are they acting on it), confidence (do we know who it is) and freshness (how recent it is).
Does the language model decide on its own?
No. The model reads each post and suggests a label; fixed rules check what it says, and a person makes the final call before anything is delivered.
How do you know which business wrote a post?
Only from what is public: a company account, a website in the bio, or the post naming the company. We never guess who is behind an anonymous account.
Can I get the same lead twice?
No. Every delivery is recorded, so a lead is never sent to you twice. The same record is what lets us offer a niche exclusively.
Does my feedback change the results?
Yes. You tell us which leads were useful, contacted or won, and that feedback tunes the scoring for your feed.
Is Kateion GDPR-compliant?
Kateion relies on legitimate interest: it uses only public posts about business matters, keeps the minimum, deletes source items after 180 days and removes anyone who asks the same working day. The privacy policy has the details.
Last updated 2026-10-06
See a week of real leads before you pay anything.
- Free for one week, then from €500 a month. Cancel any time.
- Ten checked leads in the pilot week, each with the post linked
- Tuned to the competitors and niche you name
- One question back: which leads were worth a call?