Intent data is any evidence that an account is researching what you sell. The term covers three quite different things, and most arguments about whether it works are really arguments about which of the three somebody bought.
The three sources are not comparable
- First-party intent is behaviour on surfaces you own: pricing page visits, documentation reads, repeat sessions, product usage on a free tier. You observe it directly and you know who did it, at least to the degree your forms and tracking allow.
- Second-party intent comes from somewhere a buyer went deliberately, typically a review site where they compared you against a competitor. Narrow, expensive, and the closest thing to a stated interest.
- Third-party intent is modelled from publisher networks and bidstream data across the open web. It is the cheapest per account, the largest in coverage, and the one people mean when they say intent data did not work.
What the third-party chain is actually doing
A third-party surge alert is the end of a sequence of guesses. Each step is defensible on its own and the errors compound.
| Step | What is assumed | Where it slips |
|---|---|---|
| Someone reads an article | The reading is related to a purchase. | Analysts, students, competitors and the merely curious read the same pages. |
| An IP is resolved to a company | The reader works there. | Home broadband, mobile networks, VPNs and shared offices resolve badly or not at all. |
| The content is classified to a topic | The topic maps to your category. | Classification is approximate, and categories overlap heavily in software. |
| Volume is compared to a baseline | Above-baseline reading means buying. | The baseline is the vendor's, derived from their panel, and you cannot inspect it. |
Run that chain and the output is a score for a company, not a person. Nobody has told you who inside the account was reading, whether they can buy, or whether they have a budget. You have been handed a reason to look, which is genuinely useful and is much less than what gets sold.
Three questions worth asking before you buy
Where does the underlying reading happen? A co-operative of business publishers and a bidstream feed are different products at similar prices. Ask for the sources by name and note what happens when the answer stays general.
What proportion of traffic resolves to an identified company? Resolution rate sets the ceiling on everything downstream. A vendor who has measured this will tell you; a vendor who has not will talk about coverage instead, which is a different number.
What defines a surge? If the threshold is proprietary, your prioritisation is being outsourced to a model you cannot see and cannot tune against your own closed-won data.
The first-party data nobody uses
Most teams buying third-party intent are not yet acting on the first-party intent they already collect. Repeat visits to a pricing page, a second person from the same domain arriving within a week, a trial account that invited colleagues: all observed, all attributable, all usually sitting in a tool nobody has wired to anything.
It is a smaller pool and a far stronger one, which is why it belongs in the scoring model as its own input rather than averaged in with firmographics. Buying a signal while ignoring a better one you own is a sequencing error rather than a budget error.
A signal that changes nothing is a dashboard
The common failure is not accuracy. It is that intent arrives into a process with no slot for it. The account gets flagged, the flag appears in a weekly report, and the seller carries on working the list they were already working, because nobody changed what happens when the flag fires.
Making it count means deciding in advance which action it triggers and who owns that action. In practice that is an assignment rule plus a change to sequencing, and if neither exists the data has nowhere to land. The accounts most worth a signal are the ones where somebody already has a reason to reach out, which is the arithmetic behind working warm before working wide.
When it is worth buying
When your constraint is which accounts, not what to say. If the message is working and the list is too long, prioritisation has real value. If the message is not working, a better-ordered list of people who ignore you is no improvement.
When the territory is bigger than the team can cover. Intent earns its cost by deciding what gets dropped. With a short list and enough capacity to work all of it, there is nothing for the signal to do.
When you can check it against outcomes. Keep a holdout: work some flagged accounts and some matched unflagged ones, and compare meetings booked after a quarter. Vendors will not run that test for you, and without it renewal is decided on anecdote. The same discipline applies to any bought data entering the system, where the measurable question is what changed, not what arrived.
Questions people ask
What is intent data?
Evidence that an account is researching what you sell. It comes from your own surfaces, from review sites, or modelled from publisher and bidstream networks across the web.
What are the types of intent data?
First-party, observed on properties you own; second-party, from places buyers visit deliberately such as review sites; and third-party, modelled from third-party web activity.
Is third-party intent data accurate?
It is a chain of inferences: reading, then company resolution, then topic classification, then a baseline comparison. It indicates an account worth looking at rather than a person ready to buy.
Why does intent data often fail to produce pipeline?
Usually because no downstream action changed. If the alert does not alter routing or sequencing, it becomes a report nobody acts on.
When should a team buy intent data?
When the target list is larger than the team can cover and the messaging already converts. Prioritising a list that is not working first does not help.
Or skip the search
Want the person who owns it after it ships?
If the work is a standing obligation rather than a bounded project, tell me which system you want built first. You get example profiles and a call with your pick before anything starts.
Placement runs through AYTalent. Engineers are employed and trained on the bench before they start with you.