← All posts AI & Data

Intent data: what is real and what is noise

The category promises to tell you who is in-market. Some of it does. Here is how we separate the signals worth acting on from the ones worth ignoring.

Intent data is one of the most oversold categories in go-to-market, and also genuinely useful. Both are true, and the difference between the two is almost entirely about which kind of intent you bought.

First-party: the strongest and most ignored

Your own engagement data — site visits, email opens and replies, content downloads, product usage, support tickets, past closed-lost — is the highest-quality intent signal available to you, and it is free.

It is also routinely the least used, because it sits scattered across four systems that do not talk to each other. Before you evaluate a single vendor, connect what you already have. On most audits this alone finds more actionable accounts than the intent product the client was about to buy.

Second-party: real, narrow, expensive

Review sites and comparison marketplaces know when someone researched your category on their platform. That is a genuine, specific signal — but it only covers the buyers who happen to use that platform, and it is priced accordingly.

Worth it when your category has a dominant comparison site and your ACV supports the cost. Not worth it as a general-purpose intent feed.

Third-party: where the noise lives

Aggregated content-consumption signals across a publisher network, resolved to companies by IP or identity graph. This is where scepticism should concentrate, for four reasons:

Every vendor in your category gets the same surge on the same day. A signal everyone has is not an advantage.

Behavioural signals: often better than intent products

Observable facts about a company frequently outperform purchased intent, and you can build them yourself: hiring activity, leadership changes, tooling adoption, funding, product launches, office moves.

They are unambiguous, cheap and timestamped. A company hiring three people for a function has told you something more reliable than a topic surge score ever will.

How to evaluate any intent source

Run it as a holdout test, exactly like a channel. Take accounts flagged as in-market, take a matched set that was not flagged, run identical outreach to both, and compare reply and opportunity rates.

If the flagged group does not meaningfully outperform, you have bought a random number generator with a good dashboard. Vendors will not love this test, which is itself informative.

The rule we use

Weight first-party highest, behavioural second, second-party third, third-party last and only as a tiebreaker. And never let a purchased signal alone trigger high-effort outreach — require it to co-occur with something you observed yourself.

Want this running on your pipeline?

We design and operate the system — outbound, paid and the data layer underneath. Book a call and we'll map it against your funnel.

Book a Strategy Call