Intent signals are observable actions that suggest a company is researching a purchase, such as visiting your pricing page, reading vendor comparisons on a review site or suddenly reading far more than usual about your category. When those actions are collected at scale and delivered as a feed, the result is called intent data.

Where a signal comes from decides most of what it can tell you. A visit to your own pricing page is about you. A spike in reading about "sales automation" across a network of publisher sites is about a topic. A comparison viewed on G2 sits somewhere in between. Each one is collected in a different way, and each one fails in its own way.

Teams get into trouble when they buy a feed without asking how it was built. The collection method tells you how fresh a signal is, who it describes (a company, almost never a person) and whether you'd be comfortable explaining its origin to your privacy officer.

TL;DR

• An intent signal is one observable action. Intent data is many of those actions collected and delivered as a feed. Both are evidence of buying intent, not intent itself.

• There are five main sources: your own site and product, review sites such as G2, publisher co-ops such as Bombora, bidstream data taken from ad auctions, and public company changes like hiring and funding.

• First-party signals come from a tracking script that matches IP addresses to companies. They're the most specific you'll get, but they stop at the company name and never tell you who visited, and in Europe they need a lawful basis.

• G2 reports views of your profile, your pricing, your category and comparison pages, and your plan decides which of those you receive. Bombora collects reading behavior through a tag on its members' sites and flags companies reading well above their own normal level.

• Bidstream data comes from the requests sent out in programmatic ad auctions, and the UK ICO found in 2019 that a single auction can expose a person's data to hundreds of organizations. Ask every provider in writing whether they use it.

• Every source reports companies rather than people, topics rather than vendors, and the same accounts to every buyer of the feed. Trust signals most when sources with different collection methods agree.

• Write a few fields to the company record in the CRM, let them decay, and only act on accounts that already pass your fit check.

An intent signal is one action, intent data is the feed

An intent signal is a single event you can point to: one pricing-page visit, one comparison viewed, one job ad posted for a role your product supports. Intent data is the broader category, the collection and delivery of many such events, usually as a subscription or as a field a tool writes into your CRM.

The difference matters in practice. "Intent data" sounds like something you switch on and trust. "Intent signals" keeps your attention on the individual events and whether each one means anything, and no feed can be better than the individual signals it carries.

Both sit below buying intent, which is the account's actual readiness to buy and can only be inferred. Both also sit inside the wider family of buying signals, which adds changes in fit and company events. What you observe are signals. What you infer from them is intent.

Where intent signals come from

Sorting intent signals by where they're collected tells you more than sorting them by type, because the collection method sets their limits.

The sections below go through the first four, because they're what people usually mean by intent data. Company changes are better handled as trigger events, which have their own entry.

First-party intent signals from your site, product and inbox

First-party signals happen on surfaces you control. A tracking script on your site records page views and sessions, your product logs what users actually do, and your email tools record replies. Because these actions are about you, they're the most specific intent signals you'll ever get, and nobody else sees them.

The weak point is identity. Most visitors never fill in a form, so their visits arrive anonymous. Tools close part of that gap with a reverse IP lookup, which matches the visitor's IP address to a company's network. HubSpot's buyer intent documentation describes its tracking code collecting IP addresses and other online identifiers so visits can be matched to companies, and it stores the result as company-level activity. That's typical of the category: you learn which company visited, not which person.

Matching also comes up empty more often than vendor demos suggest, because people working from home or on mobile data browse from consumer internet addresses. Turning a matched company into a named contact is a separate job for identity resolution and website visitor deanonymization, and it's the work our website visitor tracking workflow does for clients.

In Europe, first-party doesn't mean free to use. Tracking scripts can fall under the same ePrivacy rules as cookies, so settle the legal basis with whoever owns privacy before a script goes live.

Review-site signals from G2

Review sites see buyers at a useful moment, while they compare products. G2 sells this as Buyer Intent, and its data reference spells out what each signal means. A view of a category page that includes your product is described as early-stage research. A view of your profile indicates direct interest, and a view of the pricing page on your profile points to cost consideration or purchase readiness. A comparison page showing you next to at least one competitor indicates direct evaluation. There's also a competitive signal for when a buyer looks at a rival's pages in a shared category and you aren't part of it.

That built-in grading is the main appeal. The signal already tells you roughly how far along the account is, and a comparison signal tells you who else is on the list.

The limits come from the same place as the strengths. You only see buyers who research on G2, and only in categories where you're listed. G2's Buyer Intent overview also notes that not all plans receive signals and some receive more than others. Signals arrive per buyer organization, with a count of unique visitors rather than names.

How a publisher co-op like Bombora collects intent data

Third-party intent data tries to catch research that happens nowhere near your site. Bombora is the best-known example of the co-op model. Its explanation of Company Surge describes the data as coming from a cooperative of publishers it has direct relationships with, and says no bidstream data is used and that consent strings can be traced. Members run Bombora's tag on their sites, the tag records what businesses read, and the content is classified into topics.

The output isn't a raw count of page views. A company surges on a topic when it reads noticeably more about it than its own usual level. A large enterprise that always reads about cloud security doesn't light up every week, while a mid-sized firm that suddenly starts reading about it does. That baseline is the clever part, and a surge typically feeds the intent side of a model that rates fit and intent separately rather than acting as a trigger on its own.

Know what you're buying. A surge says a company is reading about a topic, not about you, and topics are broader than any one product. The step from reader to company runs through IP and identity matching like every other source, so the result is company-level. And every subscriber in your category can see the same surging accounts, which means the timing advantage is shared.

Bidstream data and the provenance question

Some intent data is built a different way. Each time a page with programmatic ads loads, an auction runs, and bid requests go out to potential buyers of that ad space. In its 2019 report on adtech and real-time bidding, the UK Information Commissioner's Office lists what most bid requests carry, including the user's IP address, cookie and user IDs, browser and device details, and location. Collect enough of those requests, map the IP addresses to companies and the pages to topics, and you have intent data without any publisher agreement at all.

The appeal is reach. The problem is provenance. The same ICO report found that one visit to a website, prompting one auction, can result in a person's personal data being seen by hundreds of organizations, and it found that the consent mechanisms the industry relied on fell short where sensitive data was involved. Bombora markets itself partly on not using bidstream, which tells you the industry knows where the fault line runs.

You don't need to become an adtech expert. Before you sign with any intent provider, ask in writing where the data comes from, whether any of it is bidstream, and what consent sits behind it. If the answers stay vague, give that data less weight.

What every intent source leaves out

The collection methods differ, but the blind spots are mostly shared:

• Companies, not people. Almost every source reports at account level. Someone still has to work out who in the company is doing the research, and whether that person matters to the decision.

• Topics, not vendors. Outside your own site and your review profile, signals tell you a subject is warm, not that anyone is considering you.

• Noise that looks like buyers. Students, consultants, agencies researching for a client and your own competitors all generate signals.

• Lag. Third-party feeds tend to be refreshed in batches, so a signal can be days old by the time a rep sees it. Your own site is the only source that can be close to live.

• Shared timing. Anything you can buy, your competitors can buy too.

None of this makes intent data useless. It means each signal should be weighed rather than counted, and three things do most of the weighing: how close the action sits to a decision, how recent it is and whether it repeats.

Corroborate across collection methods

The strongest case comes from signals collected in different ways that agree. A Bombora surge on your topic, a G2 comparison that includes you and two pricing visits from the same company in one week are three independent methods pointing at one account. Each alone could be noise. Together they're hard to explain away.

The reverse holds as well. Three surges on related topics from the same feed are less convincing than they look, because they share the same blind spots. Variety of source counts for more than volume.

Getting intent signals into the CRM without flooding reps

Intent data that sits in a vendor dashboard gets checked when someone remembers. The signals need to land where reps already work, on the company record.

In most setups we build, Clay pulls signals from whichever sources a client uses, matches them to existing accounts and writes back a handful of fields: the strongest recent signal, its source, its date and a combined intent score that decays over time. A few fields get read. Twenty columns of topic scores get ignored.

Filter on fit before anything reaches a person. Watching intent signals only on accounts your ICP scoring model already qualifies keeps reps away from busy accounts you'd never close.

What happens next depends on where the signal came from. A strong first-party signal from a good-fit account calls for a same-day, relevant touch, which is how inbound-led outbound works. A third-party surge on a high-fit account is a reason to warm the account up and watch for a first-party signal to follow, the pattern behind most signal-based selling. Put together, the sources give you a partial view of the dark funnel, the research buyers do before they ever contact you. For a longer look at the providers themselves, see our blog post on what intent data is.

Wrapping up

How much an intent signal deserves your trust depends on how it was collected. Your own site tells you about interest in you, but by company rather than by person. Review sites catch evaluation in progress, within their own audience. Publisher co-ops see category research far from your site, and bidstream data reaches furthest while raising the most questions about where it came from.

None of them hands you a buyer. What they give you is a shorter list of companies worth watching closely, and the advantage comes from combining sources that fail in different ways, checking fit first and getting the result to a rep while it still matters.

If you're deciding which sources are worth paying for, we're glad to think it through with you. We usually wire the chosen feeds into Clay and whichever CRM a client runs, set the fit filters and decay rules, and write down how each signal is weighted, so the next person to touch it knows why it works the way it does.

Frequently asked questions