Ask most sales intelligence platforms what a buying signal is and you will get a version of the same list: funding rounds, leadership changes, hiring, technology adoption, maybe some anonymised web activity. Useful signals, all of them. They have one thing in common, which is that they already exist as a field in a database somewhere.
That is exactly why every platform has them, and exactly why they are rarely a differentiator. If a signal is in a structured feed, everyone selling to your market can buy the same feed.
The signals that actually predict a purchase in your specific category usually are not in any feed. They are sitting in documents.
Why the standard taxonomy is the floor, not the ceiling
A fixed signal taxonomy is a commercial decision, not a technical limit. A platform serving thousands of customers has to pick signal types that apply broadly, because a signal relevant to forty customers cannot justify the engineering to maintain it.
So you get the universal set. Funding. Hiring. Leadership moves. These correlate loosely with buying across most of B2B, which is what makes them sellable at scale, and also what makes them blunt for you specifically.
If the signal fits in a dropdown, your competitors can buy it too.
The signals that genuinely predict deals in a specific category are narrow by nature. They matter enormously to a small number of companies and not at all to everyone else. No platform will build them for you, and that is not a criticism — it is arithmetic.
Where the good signals actually live
Regulatory registers and published judgements
Regulators publish. Enforcement outcomes, licence applications, certification lapses, inspection results, register changes. This material is public, dated, and about as authoritative as a source gets, because the regulator's own name is attached to it.
If you sell compliance, risk, safety or quality software, a published finding against a company is a stronger buying signal than any funding round. Something is now on record that somebody is accountable for fixing, usually within a defined period.
Tenders and procurement notices
A published tender is a company telling you, in writing, what it intends to buy and roughly when. It is the least ambiguous signal that exists.
It is also awkward to work with, which is why it stays underused: portals are fragmented by country and sector, formats vary, and the useful detail is usually buried in an attachment rather than in a title field. That awkwardness is precisely the moat.
Job specifications, read properly
Most platforms count job postings. The count tells you a company is growing. The text tells you what they run and what they are about to change.
A specification asking for experience with a named system tells you the system is in place. One asking a candidate to "lead the migration away from" it tells you something considerably more valuable, and with a date attached.
Company filings and annual reports
Filings are structured for the financial fields and completely unstructured everywhere else. The narrative sections — strategy, principal risks, post-balance-sheet events — are where companies say what they intend to do next, in language they have had lawyers check.
End-of-life and deprecation calendars
Software vendors publish support end dates years ahead. Anyone still running a version approaching end of life has a deadline they did not choose. Cross-referencing a deprecation calendar against detected technology turns a fixed public date into a dated, per-account reason to call.
Trade press and local reporting
Sector publications cover things the national press ignores: a site opening, a contract won, a regulatory change coming for an industry. Often this is the earliest public mention of something that becomes a buying event months later.
Why this is harder than it sounds
If this material is public, the obvious question is why everyone is not already using it.
It does not arrive as data. There is no field called "is migrating away from vendor X". There is a paragraph in a job advert that implies it. Turning that into something you can rank on means reading the document and making a judgement.
Each source behaves differently. Regulators, procurement portals and filing services each have their own structures, update rhythms and quirks. There is no single integration.
Language models will confidently invent these signals. This is the part people underestimate. Ask a model to find companies migrating off a platform and it will return a plausible, well-written list, some of which is fabricated. Generating candidate signals is easy and getting cheaper. Verifying them against a live source, with a date and a quote you can stand behind, is the actual work — and in our experience a meaningful share of raw candidates do not survive that check.
They decay at different rates. A leadership change stays relevant for a quarter or more. A tender with a closing date is worthless the day after it closes. Treating every signal with one freshness window makes half of them misleading.
What a usable signal looks like
Whatever the source, a signal is only worth ranking on if it carries four things:
- A source you can open. A link to the original document, not a citation of a provider.
- A date. When the thing happened, not when it was scraped.
- A verbatim quote. The actual sentence, so a rep can judge it and open a call with it.
- A stated reason it matters. Why this event predicts a purchase in your category, written down so it can be argued with.
A signal missing any of these is a score wearing a costume. We have written more on the difference between intent data and verifiable signals, but the practical test is simple: can a rep check it in ten seconds, and would they be comfortable quoting it on a call?
Finding yours
The useful exercise is to work backwards from deals you have already won. Not the demographics of the buyer — the trigger. What changed at that company shortly before they engaged?
Then ask whether that change leaves a public trace. Surprisingly often it does: a regulator published something, a specification appeared, a filing mentioned it, a trade publication covered it. Where the trace exists, you have a signal class nobody is selling.
Do that across ten or fifteen closed-won deals and patterns emerge fast. Usually two or three signal classes account for most of them, and they are rarely the ones in the standard dropdown.
The point
Everyone can buy funding data. Almost nobody is reading the tender that says, in plain language, what a company intends to purchase this year.
The advantage is not in the feed. It is in the documents nobody has the patience to read — and in verifying what you find there well enough that a rep will trust it.
If you want to know which signal classes predict deals in your category, book a call. Working that out is the first thing we do.