explainer

Intent signals, intent data, and what people actually say.

Published · sources checked the same day

Intent data is inferred from what a company reads. Intent signals, in the narrow sense, are things that happen to a company, such as a funding round or a new hire. Both point at accounts, and both are inferences about what the people inside them want. Stated intent is different in kind: a person writing, in their own words, that they need something. It is the rarest of the three, the most direct, and the easiest to misread.

The words are used loosely. Many vendors call all of this intent signals and split it into first-party, third-party and contextual. This page sorts it by what the evidence actually is: what a company reads, what happens to it, and what a person says.

The three, side by side

Swipe the table sideways to compare all three.

Intent dataTrigger eventsStated intent
What it isWhat people at a company read and research, aggregatedThings that happen to a companyA person writing that they need something
Where it comes fromPublisher co-ops and review sites (third-party); your own site and email (first-party)Funding news, job posts, leadership changes, tools added or droppedPosts and threads, reviews, replies, inbound email, support tickets, chat
Points atUsually an account; first-party can be a known contactAn accountA person, in their own words
Tells youAn account is researching a topic more than usualCircumstances changed in a way that often comes before a purchaseWhat they need, often with a deadline, a budget or the tool they are leaving
Cannot tell youWho is reading, or whyWhether they want your category at allAnything, until it is read correctly: most of it is noise
Best used forChoosing which accounts to watchTiming outreachDeciding whom to answer, and what to say

Intent data: what a company reads

Third-party intent data aggregates reading and research from sites a provider can see. Bombora, one of the larger providers, describes its data as showing which accounts and buying-group members are researching a topic, based on the content they consume across a consent-based co-op of B2B publishers. G2 Buyer Intent reports which companies viewed your product profile, your category, a comparison with a competitor or a list of alternatives on G2, at company level with a count of how many people. First-party intent is the same idea on your own properties: pricing-page visits, repeat visits, clicks in your emails.

Its strength is reach: it can surface an account before anyone there has spoken to you. Its limit is that it cannot say who is reading or why. Reading about help-desk software is consistent with a team about to buy one, and equally with research, a report or plain curiosity. And anyone who licenses the same data sees the same surge.

Intent signals: what happens to a company

Trigger events are facts rather than inferences: a company raised money, hired its first sales leader, opened an office, started or stopped using a tool. They are dated and usually public, which makes them good for timing, and they go stale. Clay’s guide treats a pricing-page visit as worth acting on within a day or two and a funding round within a few weeks.

What they cannot do is tell you whether the company wants what you sell. A funding round makes many purchases more likely and none of them certain.

Stated intent: buying signals in the buyer’s own words

Sales trainers have long taught reps to listen for verbal buying signals: a question about price, a question about implementation, a complaint about the current vendor. HubSpot’s guide defines buying signals as the behaviours and statements that show a prospect is considering a purchase. Those statements now happen in writing, before any call, and at a volume nobody reads by hand: a thread asking what to replace a tool with, a reply to a newsletter, a support ticket that ends with a deadline. The written forms, ranked by strength, are on buying signals.

Most intent-signal guides leave this out, or file it under social activity, one item on a long list. It deserves a category of its own for two reasons. It is the only kind where a person states the need instead of a vendor inferring it. And it is the only kind where misreading it is a public mistake: you answer someone who was joking, or pitch someone who was venting.

Why it is hard to read

We publish 158 public posts that our own scorer put forward as likely buyers and a human reviewer rejected. The reasons: 65 were too vague to act on, 61 were about something else, and 32 came from people who genuinely were shopping, some naming a shortlist, a team size or an approved budget, for a different category. Real stated intent, just not for that product. A buying signal only means something relative to what you sell.

“Our help desk renewal is in March and the quote went up again. What are teams of about 40 agents moving to?”

Buyer A current tool, a deadline, a team size, and an open question about what to switch to.

“Third sync failure this month. If it’s not fixed by Friday we’ll start looking at alternatives.”

Leaving A support ticket that states an intent to churn, with a date. The same judgement, pointed at retention.

“Anyone have tips for getting more customers?”

Too vague A real problem, but nothing specific enough to answer with a product.

“We’re choosing between two call-recording tools for a 12-person team. Budget’s approved.”

Buyer, wrong product Genuine buying intent, with a shortlist and a budget. Worth nothing unless you sell call recording.

“Oh great, another price increase. Love paying more for fewer features.”

Venting Angry, sarcastic, and no sign of leaving. A keyword on “price” fires; a reply would land badly.

“If I ever start a podcast, which host would you pick?”

Hypothetical No need yet and no timeline. Worth an answer, not a pitch.

“Has anyone tried this tool? Heard it’s great for exactly this problem.”

Possibly the vendor A question that is really an advert. The phrasing of a buyer, the motive of a seller.

Examples written for this page, modelled on the categories above. They are not real posts.

Which one you need

Judging stated intent at scale

Three ways to do it, in rising order of judgement:

  1. Keyword alerts fire when a word appears. Cheap and complete, and every post containing the word is forwarded, the jokes and the sellers included. Compared: F5Bot, Syften.
  2. One LLM prompt handles clear cases well. On ambiguous text it still returns a confident label, and nothing tells you which labels to double-check. Why three judges beat one prompt.
  3. A panel of judges with different dispositions, a Skeptic, an Analyst and an Optimist, reads borderline text independently. When they agree, you can act. When they split, you see the split, so you read exactly the cases that need a person. Clear-cut text is decided without convening the panel.

Your agent can call the third as an API or an MCP server: you send text your agent, bot or scraper already has, and get back a verdict and whether the judges agreed. SignalPipe does not sell intent data, track what companies read or decide whom to contact. You do.

See a verdict on your own text.

Paste a real post, email or ticket into the live panel and watch the three judges rule on it. Free, and no account needed.

Questions

What is the difference between intent signals and intent data?

Intent data is one kind of intent signal: the behavioural kind, aggregated from what people at a company read and research, and usually reported per account. Other intent signals are events, such as a funding round, a new sales leader or a tool being replaced. Both are inferences about an account. Neither is a person saying they want to buy, which is stated intent.

What are intent signals in sales?

Observable actions or events suggesting that an account may be moving towards a purchase: visits to your pricing page, a surge in reading about your category, a job posting for the role that would use your product, a funding round. Sales teams use them to decide which accounts to prioritise and when to reach out.

Is a Reddit post or a support ticket intent data?

Not in the sense intent-data vendors use the term. Intent data is aggregated behaviour, usually sold per account. A post or a ticket is one person’s own words about their own need: stated intent. Most intent-signal taxonomies either leave it out or list it under social activity.

What are examples of buying signals in writing?

Asking what others use to replace a named tool; stating a deadline, such as a renewal date; giving a team size or a budget; comparing two products by name; saying a current vendor has failed and alternatives are being considered. The written equivalents of the verbal buying signals sales trainers teach.

How accurate is third-party intent data?

It shows that people at an account are reading about a topic more than usual. It cannot say who they are or why, and reading about a topic is consistent with buying, research or curiosity. Anyone who licenses the same data sees the same surge. Treat it as a way to prioritise accounts, not as proof that one is buying.

Can an LLM detect buying intent in text?

On clear cases, yes. The hard cases are sarcasm, venting, hypotheticals, sellers phrasing an advert as a question, and real buyers of a different product. A single prompt still returns a confident label on those. Several independent reads, with disagreement surfaced rather than averaged away, tell you which verdicts need a person.

Does SignalPipe sell intent data?

No. SignalPipe does not track what companies read, identify website visitors or sell account lists. It judges text you already have, such as posts, emails and tickets, for whether the author wants to buy what you sell, and says when its three judges disagree.

Sources