How buying-signal scoring actually works

“Intent data” is the most oversold term in go-to-market software, largely because a score is unfalsifiable unless you can trace it. This explains what buying signals really are, the three sources they come from, how scoring is constructed, and the specific question that separates a real signal from a number a vendor generated to look clever.

What counts as a signal

A buying signal is any observable event correlated with a company becoming ready to purchase. They fall into three groups.

Buying signals — capacity to spend

Funding rounds, new investment, acquisitions, cost-cutting programmes, competitor churn, tech-stack changes. These tell you money is moving.

Timing signals — the problem is live now

Job specs describing the pain you solve, department expansion, new offices, product launches, pricing-page visits, category research on review sites. These are the highest-conviction group, because a company hiring four people to hand-fix a process has already priced the problem and put it in a budget.

Person signals — someone new can say yes

Executive appointments, role changes, public posts, conference talks, a champion arriving or leaving. New executives buy: they have a mandate and a window in which to use it.

Where the data comes from

Most tools sell you the third row while implying the fourth. The distinction is worth understanding: co-op intent tells you someone at a company like this read something like this. Live web evidence tells you this company published this sentence on this page on this date.

How a score gets built

A defensible score has four components, and you should be able to see all four.

1. Weight. Each signal type contributes a stated amount. A pain-matched job spec should outweigh a single pricing-page visit; if it does not, ask why.
2. Decay. Signals expire. A funding round from fourteen months ago is history, not intent. Every signal needs a stated half-life.
3. Stacking. Signals that co-occur are worth more than the sum of their parts. A new COO plus four hires in the function that COO owns is a different proposition from either alone.
4. Fit multiplier. Intent without fit is noise. A perfect signal at a company that will never buy from you should score low.

Why source citation changes how reps work

An uncited score can only be used one way: to sort a list. A cited signal can be used in the email itself, on the call, and in the internal case for prioritising the account.

Compare the two openers this produces:

“I noticed your company might be exploring solutions in this space”

“You’ve posted four compliance-ops roles this month, all mentioning reducing account-opening turnaround from nine days — that’s the specific number we move.”

The second is only writable because the underlying evidence was retained rather than compressed into a score. That is the practical argument for citation: it is not about auditability for its own sake, it is that the evidence is the raw material of the email.

Stacking, and the “why now” narrative

Single signals are weak. The value appears when several land on one account inside a short window, because that is what a real buying process looks like from the outside: a new executive, a hiring push in their function, and a visit to your pricing page are three views of the same internal decision. A system that only reports signals leaves you to assemble that story. A better one synthesises it — and can show you the three sources it was built from.

What signals cannot tell you

They cannot tell you budget is approved, that your champion has authority, or that the timing is anything other than plausible. Signals reorder your queue and sharpen your first line. They do not qualify a deal, and a rep who treats a high score as qualification will have a bad quarter.

← All guides · How to evaluate AI SDR alternatives · Cold email deliverability