B2B Intent Data and Lead Scoring on a Small Team

Search b2b intent data and read the first ten organic results. Every one of them is published by a company that sells intent data. That is not a conspiracy. It is what happens when the only people with a content budget in a category are the vendors in it. The AI Overview sitting on top of those results cites the same vendors.

So here is the review nobody on that page will write. You already own three intent signals, for free, right now. Most B2B teams under ten people should run on those for a year before they spend a dollar on a fourth.

And the scoring model that decides which of those signals matter should be built backward from the deals you already closed — not forward from a point table you invented on a Tuesday afternoon.

What is intent data actually telling you that you do not already know?

Intent data is a probability estimate, not a fact. G2's own Buyer Intent Data Providers category page defines buyer intent as 'the likelihood that a customer will purchase a product' and lists 95 providers selling you that likelihood as of July 2026.

The split that matters is first-party versus third-party. G2 draws it plainly: 'first-party data alone is primarily used for visitor identification software. What sets buyer intent data apart is its integration of first-party data with second or third-party sources.' Translation — the thing you pay for is the outside data, and the outside data is the part you cannot verify.

How is that outside data captured? G2 again: 'Most intent data is captured through content consumption, like website visitors, their engagement with an email, or even social media.' A publisher co-op sees anonymous reading behavior, matches it to a company by IP or cookie, and flags a topic surge. Bombora, ZoomInfo, 6sense, Demandbase and Cognism all sell some version of that.

Now read what a topic surge actually tells you: somebody, at a company you may or may not sell to, read something about your category recently. Not who. Not whether they have budget. Not whether they are a buyer, an intern, or a competitor doing research on you.

Gartner's B2B buying research says 75% of B2B buyers prefer a rep-free sales experience, and that 99% of B2B purchases are driven by organizational changes. A surge signal cannot see the reorg, the new VP, or the budget cycle that is actually causing the purchase. It sees the reading, weeks after the decision to look started.

Which intent signals do you already own for free?

Three, and all three are first-party, meaning nobody modeled them for you: your Google Search Console query report, the behavior of returning visitors on your own site, and your profile activity on the review platform your category buys from.

Start with Search Console. Google's Performance report documentation describes the queries dimension as grouping 'your data by the search query users typed,' alongside four metrics: clicks, impressions, CTR, and average position. That query list is the only place you get a buyer's own words. Somebody typing 'alternatives to [your competitor]' or '[your category] pricing' and landing on your site is stating intent out loud. No vendor model required.

Two honest caveats. The default view is the last three months, so pull and store it monthly or you lose history. And Google warns that 'even if a query appears in your list, you might not see your site in results if you run the same query in Google Search' — this is aggregate, not per-person. Use it to spot which problems your market is shopping for, not to identify a lead. Our Search Console playbook for service businesses covers the pulls that matter.

The second signal is what people do on your own site. A visitor who hits your pricing page, reads one case study, and comes back within seven days is telling you more than any purchased topic surge. You already have that in your analytics. Nobody sold it to you because nobody can.

The third is the review platform. Buyers do their comparison shopping on G2, Capterra and Clutch — which is exactly why G2 turns around and sells that activity back to vendors as an intent product, listed in its own buyer-intent category. Being the profile a buyer opens is a purchase signal. Owning that listing is cheaper than renting the data about it. We wrote the full approach in review-platform strategy for G2, Capterra and Clutch.

Is a third-party intent data subscription worth it for a small team?

For a team under about ten people, almost never — and the clearest evidence is on the vendors' own G2 listings, where 'Inaccurate Data' is the single most-tagged complaint on ZoomInfo's GTM Workspace (205 mentions) and on 6sense Sales Intelligence (131 mentions).

That is the category's own users, on the category's own review site, naming accuracy as the top problem with the exact thing they bought. Here is how five of the best-known listings in that category present themselves as of July 2026.

Verdict: skip it. Not because the products are frauds — at enterprise scale, with a twelve-person SDR bench burning through accounts, a probability ranking of 4,000 target companies pays for itself. Buy it when you have more sales capacity than qualified leads. A ten-person firm has the opposite problem: more leads than time to call them. Buying a bigger list does not fix that. Scoring the list you have does.

Notice the pricing column, too. Three of the five listings above will not publish a price. We think 'get a quote' is a red flag in any category, including ours — which is why our own pricing is on the site.

How do you build a lead score backward from your closed-won deals?

Export your last 20 closed-won deals and your last 20 closed-lost, put them side by side, and score only the attributes that visibly differ between the two lists. Twenty of each is enough to see the obvious splits. It is not statistically rigorous and it does not need to be — it beats any point table you invented from nothing.

Every vendor guide runs this the wrong way round. They hand you a template — visited pricing page, +10; opened three emails, +5 — with numbers that came from nowhere. Yours came from your actual revenue.

The mechanics take an afternoon. Pull both lists into one sheet, then for each attribute ask one question: does it appear more often in the won column than the lost column? If yes, it is a positive. If it appears equally in both, it is noise, and noise in a scoring model is worse than nothing because it launders bad leads into good scores.

• Pull the last 20 closed-won and 20 closed-lost deals out of your CRM, with the source, company size, industry, job title of the first contact, and the pages they hit before the form fill.
• Split every attribute into two buckets: fit (who they are — size, industry, title, region) and engagement (what they did — pages, return visits, the query they arrived on).
• Keep only attributes where the won and lost lists actually diverge. Most will not. That is the finding, not a failure.
• Weight fit heavier than engagement. A perfect-fit company that read one page beats a bad-fit company that read nine.
• Backtest it: run the score against the 40 deals you already know the outcome of. If your model would have ranked the losses above the wins, it is wrong. Change it before it touches a live lead.

That backtest is the entire trick, and it is the step that separates a scoring model from a horoscope. Your CRM already holds the data — HubSpot, Pipedrive or Salesforce will all export it to CSV in about a minute.

Where does the score threshold go, and who decides it?

Sales decides, not marketing, and the correct threshold is the one that produces about as many MQLs per week as your reps can actually call within an hour of the form fill.

That hour is not arbitrary. In a 2011 Harvard Business Review study, firms that contacted an online lead within an hour were nearly seven times as likely to qualify that lead — defined as having a meaningful conversation with a key decision maker — as firms that waited just one hour longer, and more than 60 times as likely as firms that waited 24 hours or more. The same research audited 2,241 US companies and found the average response time to a web lead was 42 hours, with 23% never responding at all.

Read that against your threshold. If your model hands sales 60 MQLs a week and sales can only call 25 of them inside an hour, your threshold is too low and your scoring model is actively destroying pipeline. Raise it until the numbers match capacity. We go deeper on this in speed to lead.

Then measure sales acceptance rate — the share of MQLs your reps agree were worth calling. Ignore the benchmark numbers floating around the internet; almost none of them name a dataset. Measure your own baseline for a month and manage the trend. If acceptance falls, the threshold moved or the model drifted. That is the only number in this whole system worth arguing about in a weekly meeting.

What should score negative, and why does nobody write those rules down?

Negative scoring is the highest-leverage half of the model, and roughly nobody publishes it, because a vendor's commercial incentive is to make your lead count go up — not down. Yours is the opposite. A small sales team drowns in volume, not in scarcity.

These are the rules that quietly protect a two-rep team's calendar. Write them once, in the same sheet as the positive rules, and they will save you more hours than any tool you buy this year.

• Job-seeker titles — anyone whose title contains 'student', 'intern', 'seeking', 'looking for work', or who arrived via your careers page. Auto-disqualify, never route to sales.
• Competitor domains — maintain the list by hand, add to it every time you lose a deal. Competitors read your pricing page more carefully than your buyers do.
• Agency and reseller domains, unless partnerships are a real channel for you. If they are, route them somewhere else, not to a closing rep.
• Existing customers filling in a top-of-funnel form. This happens constantly and it inflates lead counts in every report you will ever read.
• Out-of-region — if you only serve the US, a lead in another country is a zero, not a low score. Zeros and low scores behave differently in a ranked list.
• Free-email domains, with a caveat: gmail addresses are a genuine negative in enterprise B2B and a false negative for small businesses, where the owner really does use gmail. Test it against your own closed-won list before you turn it on.
• Content-only behavior — someone who has read six blog posts and never once touched a pricing, service, or contact page. High engagement, zero purchase intent. This is the lead a naive engagement score ranks first.

That last one is worth sitting with. An engagement-weighted model with no negative rules will reliably promote your most enthusiastic non-buyer to the top of the queue, and your rep will call them first.

How do you run all of this in a spreadsheet before buying software?

One sheet, five columns, and a twenty-minute weekly review. That is the whole system, and it will out-perform a platform you bought but never configured — which describes most marketing-automation seats sold to companies of this size.

The five columns: lead, fit score, engagement score, negative flags, total. Sort descending. Reps work top-down. Anything with a negative flag goes to a second tab and nobody calls it.

Add one rule most people forget — decay. An engagement score earned six weeks ago is not the same as one earned yesterday. Halve the engagement component after 30 days of silence, zero it after 90. Fit does not decay; a good-fit company is still a good-fit company next quarter. Only behavior goes stale.

The weekly ritual: sales and marketing look at the leads that scored above the threshold and did not convert, and the ones that scored below and converted anyway. The second list is where the model learns. It is the same feedback loop that makes an organic program work, which is why we run scoring and search together inside our B2B SEO service rather than treating them as separate departments.

Gartner's research adds a useful pressure test here: B2B buyers are 1.8 times more likely to complete a high-quality deal when they engage with supplier-provided digital tools alongside a sales rep, rather than independently. Your scoring model exists to decide where that rep spends their hours. It is not a reporting artifact.

When is it finally time to buy the tool?

Buy when the spreadsheet breaks for a reason you can say out loud in one sentence — usually somewhere north of about 50 scored leads a week, when the manual review starts costing more than a software seat.

Three legitimate triggers, and none of them is 'a competitor has one':

• Volume — the weekly review takes more than an hour, every week, for a month. That is a real cost. Automate it.
• Routing — you have more than two reps and territory or specialization rules that a human keeps getting wrong.
• Sales capacity exceeds lead supply — you have reps sitting idle. This, and only this, is when third-party intent data starts to make sense, because now you genuinely need to manufacture more accounts to call.

Until one of those is true, the tool is a purchase, not a solution. And when you do buy, buy month-to-month if the vendor allows it, and set a kill date — 90 days, no qualified leads attributable to the tool, cancel it. We apply the same rule to our own channels, and it is the only honest way to hold a spend accountable.

The uncomfortable truth in this whole category is that the scoring model does most of the work and costs nothing, while the data subscription does less of the work and costs five figures — and the second one has all the marketing behind it. That is worth remembering the next time an AI Overview confidently recommends a vendor to you.

If you want the search half of this done properly — the query data, the pages that pull in real buying intent, the organic pipeline that feeds the score — that is what our B2B SEO service does. Get my free audit and we will tell you which of your pages are already attracting buyers and which are just attracting readers.