Buyer Intent Signals: Which Ones Actually Correlate With Closed Deals

A separate article on this site walks through what buyer intent data actually is, where the signals come from, and how to build a scoring workflow around them. This one skips the primer and asks a narrower question: of all the signals vendors sell, which ones actually show up more often in deals that close, and which ones just make a dashboard look busy?

The answer matters because most teams buy an intent tool, wire it into their scoring model, and never go back to check whether the signal they are paying for actually correlates with anything. It usually does not get checked because checking requires closed-loop data that most CRMs are not set up to produce automatically.

The signals that hold up

Two categories consistently show a real relationship with closed revenue when teams actually run the analysis.

Product usage and expansion signals, for companies with an existing customer base or a free tier, are the strongest predictor of anything in this space. A workspace adding seats, hitting a plan limit, or turning on a feature associated with expansion is not an inference about intent, it is intent, recorded directly. These signals correlate with expansion revenue at a rate that dwarfs anything third-party data offers, because there is no matching problem and no ambiguity about who is behind the behaviour.

First-party web behaviour on high-intent pages is the second reliable category. Repeated visits to a pricing page from more than one person at the same account, in a short window, correlates with deals that close faster and at a higher rate than average. This is not a subtle effect. Teams that segment their closed-won data by "multiple pricing-page visitors in the two weeks before the deal was created" versus everyone else almost always find a meaningful gap in win rate.

The signals that look better than they perform

Generic third-party topic surges are the category most likely to disappoint under scrutiny. A vendor reporting that an account is "surging" on a broad topic like "sales enablement" is describing content consumption across a publisher network, matched to a company with real but imperfect confidence. When teams tag opportunities sourced from these surges and compare them against opportunities with no intent signal at all, the win-rate difference is often small enough to be noise, especially once the accounts are filtered down to the ones that already fit the ideal customer profile.

Job postings are a special case worth separating out. They correlate well with timing, a company hiring for a role your product supports is a genuinely useful reason to reach out this quarter rather than next, but they correlate poorly with close rate on their own. A job posting tells you a problem might exist. It does not tell you whether this specific account will buy, has budget, or is even the kind of account that converts for you historically.

The trap: correlation with fit, not with intent

The mistake that inflates almost every intent vendor's case study is failing to control for account quality. If a scoring model already weights firmographic fit heavily, and a signal happens to fire more often on accounts that already fit the ICP well, the signal will look predictive purely because good-fit accounts close more often regardless of the signal. The signal gets credit for a correlation it did not cause.

The only way to catch this is to compare the signal's performance within a fit-controlled group: same segment, same company size band, same existing pipeline stage. If the signal still shows a gap inside that narrower comparison, it is doing real work. If the gap disappears, the signal was riding on fit the whole time.

How to actually run the test

1. Add a field to every opportunity capturing the signal, or signal combination, that prompted the outreach, including "none" as a valid value.
2. Wait for a full cycle, or two, of closed-won and closed-lost outcomes.
3. Compare win rate and average cycle length by cohort, holding segment and company size roughly constant.
4. Re-run the comparison quarterly. Signal reliability drifts as your ICP, competitive landscape and product change.

What this means for how you score accounts

Weight product usage and repeated first-party behaviour heavily. Treat job postings and firmographic events as timing triggers, not scoring inputs that should move an account to the top of a list by themselves. Treat third-party topic data as the lowest-confidence layer in the stack, useful for casting a wide net, not for deciding who gets a call this week. And run the cohort comparison at least once before renewing any intent contract, because the vendor selling you the signal has no incentive to run it for you.