First-party intent is behaviour on property you own. Third-party intent is behaviour an aggregator observed elsewhere and sold you. The useful axis is not ownership though: it is whether the data names a person or guesses at an account, and how stale it is by the time you see it.
Disclosure: Signl sells a fourth thing, public engagement data, which is person-level and is described as its own category below rather than smuggled into one of the other two. We sell no first-party or third-party intent product, so the comparison between those two has no horse in it.
The standard two, and what they actually see
| First-party | Third-party | |
|---|---|---|
| Source | Your site, emails, product, events | An aggregator's publisher and site network |
| Identifies | Usually a named person, via a cookie, a form or a login | Usually an account, inferred |
| Lag | Minutes to hours | Commonly days to a fortnight |
| Volume | Low. Only people who found you | High. Your whole addressable market |
| Cost | Effectively free. You already own the systems | A subscription, often a significant one |
| Transparency | Total. You know exactly what the event was | Limited. Methodology is usually proprietary |
| Best at | Telling you who is close to buying now | Telling you which accounts to work at all |
| Worst at | Finding anyone who has not met you | Telling you which human to contact |
The anonymity problem in the right-hand column is the one that catches teams out. An account-level score says "someone at Acme is researching your category". It does not say who, and a 400-person company has a lot of someones. Acting on it means guessing at a contact, which puts you back in cold outbound with an extra subscription.
The lag problem is the other. A surge reported a fortnight after the research happened may be describing a decision that has already been made.
The fourth category
Public engagement does not fit either definition, and forcing it into one of them is why it gets misunderstood.
It is not first-party: the conversation happened somewhere you do not own. It is not third-party as that term is normally used: nothing is aggregated, inferred or anonymised, and no panel is involved.
What it is: a named individual, in public, in their own words, at the time they said it.
| First-party | Third-party | Public engagement | |
|---|---|---|---|
| Named person | Usually | Rarely | Always |
| Their own words | Rarely | Never | Always |
| Lag | Minutes | Days to weeks | Hours |
| Coverage of your market | Only people who found you | Broad | Only people who post |
| Who else can see it | Nobody | Every subscriber | Everyone, including competitors |
| Explains why | Sometimes | No | Yes, that is the whole content |
Two of those rows are disadvantages and worth stating plainly. Coverage is poor: most of your market never writes anything in public, so this sees a minority. And it is not exclusive: a public comment is public, so a competitor watching the same source sees the same person. First-party data is the only one of the four nobody else has.
Combining them without buying everything
A sensible order, cheapest and highest-confidence first.
1. Instrument what you already own. Pricing page visits, repeat sessions, email replies, product usage for existing accounts, webinar attendance. Most teams have this data in three systems and have joined none of them. It is the strongest signal available and the only one that costs nothing per record.
2. Add public engagement if your buyers talk. Competitor threads, creators your market reads, topic searches. This covers the gap first-party leaves, which is everyone who has not found you yet. Whether it works for you is a question about your market, not about the tooling: if nobody in your niche posts, skip it.
3. Add third-party last, and only with a reason. It earns its subscription when you have a defined account list, large deals and a team that can work accounts rather than people. At small deal sizes the cost per useful action is usually worse than the alternatives.
A worked combination that does not require all three: first-party tells you Acme's VP of Finance read the pricing page twice. Public engagement tells you their controller complained about reconciliation on a competitor's post last week. Neither alone is a buying committee; together they are two named people, a stated problem and a stage.
Reading an intent vendor honestly
Five questions, in the order they expose the most.
- Does this identify a person or an account? If an account, what is your plan for choosing a contact?
- What is the median lag between the behaviour and my seeing it? Not best case.
- What was the actual observed event? "Surge" is a model output, not an event.
- What is the coverage of my specific segment? Aggregate panel size tells you nothing about your niche.
- What happens when the model is wrong, and can I see why it fired?
Question three separates measured data from inferred data, and question five is the one that distinguishes a tool you can debug from a black box. We ask ourselves the same thing: every person Signl sets aside keeps a reason in words, because a verdict you cannot inspect is a verdict you cannot improve.
Limits worth knowing
- No intent data tells you about budget or authority. All four categories describe interest.
- Account-level and person-level are not substitutes. Converting the first into the second is guesswork, and it is the step where most intent programmes quietly fail.
- Freshness decays differently per type. A pricing page visit is worth acting on today. A third-party surge may already be historical. A public comment sits between the two.
- All of it is personal data. Lawful basis, transparency obligations and the right to object apply regardless of which category it came from or how public it looked; see GDPR and B2B prospecting. This is not legal advice.
If you are weighing the operating model rather than the data, that is intent-based outbound vs cold outreach. For the strength of individual public signals, LinkedIn buying signals has the ladder, and our company-level signal types are documented in the buying signals cluster.
