Why There Is No Single Best Intent Data Provider for In-Market B2B Buyers
Intent data is a reading of research behaviour. Each provider reads a different set of places where that research happens, so the question is not which provider is best. It is which provider can see the places your buyers actually go.
A company selling infrastructure software to IT teams has buyers who read technical publications and compare vendors on software review sites. A company selling industrial equipment to plant managers has buyers who may never touch either. The same feed can be excellent for the first and close to empty for the second, and both will be shown a dashboard that looks full.
There is a second, more uncomfortable filter. Signal-based marketing is an expensive discipline that suits a narrow set of businesses. It pays when three things are true:
- A finite, nameable market. If the companies that could ever buy from you number in the hundreds, knowing which forty are moving this quarter changes what a small sales team does with its week. If the market is everybody, the same information is noise.
- A long cycle with a visible trigger. A contract renewal, a funding round, a new site, a regulatory date. When the trigger exists and is published, the timing problem can be solved, and solving it is worth a lot.
- Somebody available to act. The value is all in how fast a person responds. If your sales effort is one person already at capacity, a timing advantage converts into nothing, and the budget is better spent on capacity.
If none of the three holds, no provider is the right one. The money is better spent on being easy to find when those companies start looking.
The Four Kinds of Feed, and What Each One Can See
Vendor names change, merge and rebrand. The categories underneath them are stable, and choosing the category first narrows a long vendor list to two or three worth a trial.
| Kind | Where the signal comes from | Strong when | Blind to |
|---|---|---|---|
| Publisher co-op | Content consumption across a shared network of B2B websites, scored against each company's own normal level of research. Bombora's Company Surge is the best-known example. | Your buyers read widely on topics you can name, and you want broad account coverage. | Buyers who research off the network, and niches with few publications. |
| Review-site intent | Activity on software review and comparison sites. G2 Buyer Intent reports which companies are researching your category, your profile and your competitors. | You sell software and buyers compare vendors on review sites late in the decision. | Anything that is not bought through a review site. It reports companies, not people. |
| Publisher first-party | Readers of one media company's own sites, often down to the individual. TechTarget's Priority Engine works this way across its network of technology publications. | Your buyers are technology buyers who read that publisher. | Every buyer outside that publisher's audience. |
| Event and change signals | Public events: job postings, funding announcements, leadership moves, new locations, technology changes. | Your deals follow a visible trigger, such as a hire, a raise or a move. | The quiet research that happens before any public event. |
Account-based platforms such as 6sense and Demandbase sit on top of these. They combine several signal sources with account identification, scoring and advertising in one product. That is convenient if you will use the whole platform and expensive if you only wanted the signal.
Let your last ten deals pick the best intent data provider for in-market B2B buyers
Before any demo, pull your last ten closed deals and ask each account manager one question: where was this buyer before they talked to us? Reading analyst content, comparing vendors on a review site, hiring for a new role, opening a new office? The pattern in those ten answers tells you which category can see your buyers. It is a better guide than any vendor comparison chart, because it is about your market and nobody else's.
Most Intent Data Is Wrong About Most Accounts
Not useless. Wrong at a rate nobody measures, which is a different and more expensive problem.
Third-party intent works by inference, in three steps. Somebody at a company read something on a site in the network. The visit is matched to a company, usually through its network address. The company is then reported as researching your category. Each step is a probability, and the errors multiply. A researcher working from home may not be matched at all. A student may be counted as a buyer. One person's reading may be reported as a whole account's interest.
None of that makes the data worthless. It makes it a prioritisation tool rather than a trigger. It is useful for deciding which 50 of 500 accounts get attention this month. It is dangerous as a reason to contact somebody who has shown no interest in you in particular.
Directional, not qualifying
Intent data is good at telling you an account is researching a category. It is poor at telling you that account will buy from you. Used to decide where to look, it earns its fee. Used to decide who is qualified, it sends sales after accounts that were never going to buy, and the team stops trusting the alerts within a quarter.
Start With the Signal You Already Own
Before buying somebody else's inference, use the behaviour you can see directly. A named contact coming back three times in a week. A pricing page read twice. A document downloaded and then forwarded inside the same company. These are first-party signals. They are accurate, and they cost nothing extra. Most businesses capture none of them because nobody connected the website to the CRM.
The second source is the enquiry itself. What someone searched before they arrived, which page they landed on, and what they asked about first describe the problem better than any firmographic score. All three are thrown away by default.
Third-party data earns its place in the gap: accounts that match your profile and have shown you nothing yet. That is a genuinely useful list, and it is much smaller than the one most feeds deliver. So the order is first-party, then enquiry context, then bought data. The budget should follow the same order. In most companies it runs the other way round.
The Holdout Test for the Best Intent Data Provider for In-Market B2B Buyers
This is the test almost nobody runs, and it decides everything else. It needs records you already have and about an afternoon.
- Take last quarter's flagged accounts. Every account the feed marked as in-market or surging on your topics.
- Take a random sample of unflagged accounts from the same target list. Same size, same profile, picked at random rather than by hand.
- Count conversations in both groups. Not clicks or email opens. Replies, meetings booked, opportunities opened.
- Compare the two rates. If the flagged group converts clearly better, the feed is earning its fee. If the two numbers are close, you are paying for a list that does what a random draw would have done.
If you are buying for the first time, run the same comparison as a trial. Most providers will run your target account list against their data for a limited period. Ask for that, and agree the holdout group before the trial starts, so nobody can pick the winners afterwards. Run two providers from different categories against the same list if you can. Their overlap is interesting. Where they disagree is more useful.
Run it again before every renewal. It is the only evidence either side of that conversation can bring, and it turns "the platform feels useful" into a number.
What to Ask the Best Intent Data Provider for In-Market B2B Buyers Before You Sign
A demo shows what the product does on a good day. These questions show what it will do on your list.
- How many of our target accounts do you see at all? Send the list and ask for a match rate. A provider that recognises a third of your accounts can only ever help with that third.
- Which topics map to what we sell? Intent is reported against a fixed list of topics. If none of them describes your category closely, you are reading a neighbour's signal.
- How is a surge calculated, and how often is it refreshed? A spike measured against the account's own normal level means something different from raw volume, and weekly data on a signal that decays in days arrives late.
- Company level or person level? Account-level signals tell you where to look. Person-level data carries more privacy obligations and is worth checking with whoever owns compliance on your side.
- What does it connect to? If the alert cannot reach the right rep inside your CRM with context attached, it becomes a dashboard nobody opens.
- Will you support a holdout trial? A provider confident in its data will. One that will only show you its own success stories has told you something.
Four Things to Settle Before You Buy Any Feed
None of these is about data. All four are about what the business does with it, which is where every signal programme actually succeeds or fails.
1. Who acts, and how quickly
A signal has a half-life measured in days. If the route from a flagged account to a human conversation runs through a weekly meeting, you are buying information you cannot use. Decide the response time before you decide the provider.
2. What the first message says
"We noticed you were researching" reads badly, and buyers notice. A message about the problem itself, sent at the right moment, does not. Write the message for each signal type before the first alert fires.
3. What happens to accounts that never convert
Most flagged accounts will not become customers this quarter. Without a rule for releasing them, the list grows until the same companies are contacted again and again by people who do not know it already happened.
4. What would make you stop
Write it down at the start, in one sentence: the cost per conversation at which this is no longer worth running. If it is agreed afterwards, the decision goes to whoever argues hardest at renewal.
What to Report Once It Is Running
Intent programmes are easy to report well and hard to report honestly. Read these quarterly, and ask of each source whether it produced meetings or only activity.
| Measure | What it tells you |
|---|---|
| In-market accounts identified | How much of your target list the feed is surfacing. |
| Signal-to-engagement rate | The share of flagged accounts that replied, met or engaged. The number the holdout test is built on. |
| Response time on high-intent accounts | Whether people are acting while the signal still means something. |
| Pipeline from signal-triggered plays | Opportunities that started with a signal, kept separate from everything else. |
| Coverage of the target list | How many of the accounts you care about the feed can see at all. |
| False-positive rate | Flagged accounts that, on inspection, were never in the market. The number vendors least like to discuss. |
Two reporting habits keep this honest. Label platform-reported numbers as platform-reported, and state the attribution window on every view. And never add up credit from several platforms and present the total as revenue. Each one will claim the same deal.


