Intent Score 101: Does a High Score Mean Real Interest?

Here is the uncomfortable truth most scoring vendors will not tell you: an intent score is a hypothesis, not a verdict. A number like 87 out of 100 feels like proof, so reps treat it as permission to act. The real question underneath every score is simpler and harder: is a prospect actually interested, or does the model just think so?

I run marketing at Storylane, and I have watched good reps burn a week chasing a "hot" account that was a competitor doing research. So this piece takes a side: a high intent score is a prompt to verify, not a green light.

The strongest way to verify is engagement depth: what a prospect actually explored inside your demo and product, a signal almost every scoring model quietly ignores. By the end you will have a five-point framework to separate genuine interest from noise before you spend a single rep hour.

What Is an Intent Score, Really?

Most teams inherit their scoring model from a tool and never interrogate what the number means. That is the first mistake. A score is a compression of many signals into one digit, and compression always loses information.

Definition: An intent score is a single numeric estimate, usually on a 0 to 100 scale, of how likely an account or contact is to be actively evaluating a purchase, calculated from behavioral, firmographic, and third-party signals.

The 0 to 100 convention exists because sales teams need a shorthand to triage, not because interest is truly quantifiable to two significant figures. Most models treat somewhere in the 60 to 75 band as "sales-ready," and everything below that as nurture. That threshold is a convention, not a law of physics, and treating it as one is how teams end up trusting a number they have never pressure-tested.

The point of a score is to distinguish a casual browser from a serious buyer so reps spend time where it converts, and that is a good goal. The problem starts when the score becomes the only thing anyone looks at, which is why I am a fan of using purchase intent to prioritize outreach rather than obeying it. A score should start a conversation with your data, not end it.

Intent Score vs. Lead Score vs. Buyer Intent Data

Buyers on this topic conflate three terms constantly, and the confusion leads to bad routing decisions. Here is the clean separation I use with my own team.

The distinction that matters most: buyer intent data is an ingredient, while the two scores are recipes built from it. Confusing the ingredient for the finished dish is how teams end up acting on a signal they have not verified. Keep them separate and your model gets easier to debug.

How Intent Scores Are Actually Calculated

Under the hood, a scoring model is a weighted sum with an expiry date. Four families of signal usually feed it: behavioral (page views, demo opens, email clicks), firmographic (industry, company size, region), engagement (depth and recency of product interaction), and third-party (category research picked up off your own site). Each signal gets a weight, and most decent models apply temporal decay so a spike from last month does not count the same as one from this week.

The mechanics that separate a real model from a vanity number are decay, thresholds, and recalculation cadence. Decay means points expire on a schedule, and thresholds define the tier boundaries.

Cadence is how often the whole thing recomputes, and a model that only updates in a nightly batch will always be a step behind the buyer. A demo request is one of the strongest intent signals you can feed a model, which is why it usually carries a heavy weight.

Here is a stripped-down worked example so the abstraction gets concrete.

Add up a prospect who visited pricing (+15), opened a demo (+20), came back to view it again (+25), and matches your ICP title (+10), and you land at 70. Fold in a webinar from three weeks ago that has partially decayed to +2, and you are at roughly 72 out of 100.

On paper that is a hot lead. Whether it is a real one depends entirely on the checks in the next two sections, because the same 72 can hide either your best deal of the quarter or a well-disguised tire-kicker.

Why a High Intent Score Doesn't Always Mean Real Interest

This is the section the whole keyword hangs on, and almost nobody writes it honestly. A high intent score is a claim your model is making about a human being, and claims can be wrong in two directions. It can be fooled into a false positive, or it can miss a real buyer entirely as a false negative.

Both errors cost you, but they cost you differently. A false positive wastes rep time and pollutes your pipeline with records nobody qualified, while a false negative is quieter and more expensive: a ready-to-buy account that never lit up your dashboard and quietly bought elsewhere. I would rather over-invest in catching both than trust a single number, and the two lists below are where I start.

False Positives: When the Score Is Fooled

A false positive is a high score attached to someone who will never buy. They are more common than most dashboards admit, and they share a few recurring shapes.

• Competitor research disguised as prospect visits: your rivals study your pricing and demo more thoroughly than most customers, and every click inflates their score.
• Students, analysts, and job seekers researching for a paper, a report, or an interview, generating textbook behavioral signals with zero purchase intent.
• Bot and aggregate traffic that trips behavioral triggers at scale, especially on public pricing and demo pages.
• Procurement running an early-stage vendor scan across a dozen tools, most of which will be eliminated in the first round and never convert.

False Negatives: Real Buyers the Score Misses

The false negative is the error that keeps me up at night, because you cannot chase a signal you never saw. These are real buyers doing real evaluation somewhere your tracking cannot reach.

• Dark social research: the buyer forms an opinion in Slack communities, podcasts, and peer DMs long before they touch a trackable channel.
• Private or incognito evaluation, where the most serious buyers deliberately avoid tipping their hand until they are ready to talk.
• Offline champions who advocate for you internally but never personally visit your instrumented pages.
• Evaluators using a colleague's login or a shared device, so the activity attributes to the wrong contact or no contact at all.

The lead-leakage problem is not hypothetical. As one buyer put it, the anxiety is about the records you cannot see at all.

"We need to understand all the leads that we didn't know existed."

[product marketing, veterinary/animal-health software]

A 5-Point Framework for Verifying a Prospect Is Actually Interested

This is the part I actually want you to steal. Run these five checks before any rep acts on a high score, and you will cut wasted outreach dramatically while catching interest your model missed. Treat it as a checklist, not a suggestion.

1. Corroborate with a second, independent signal type. One signal is an anecdote, two independent signals are a pattern. A pricing visit plus a repeat demo view means far more than ten pricing visits alone, because the second signal is drawn from a different behavior and is much harder to fake accidentally. Never act on a single-signal spike.
2. Check the timing window. Intent decays fast, and a score that does not account for recency will lie to you. A 30-day-old spike is not the same as this week's activity, even if the raw number looks identical. Before you reach out, ask when the signal fired, not just how big it was.
3. Verify ICP fit before acting. A perfect behavioral score on a bad-fit account is still a bad lead, and no amount of clicking changes that. Screen for firmographics and buying-group shape first, because complex B2B purchases involve a whole committee rather than one enthusiastic contact, so a single champion at a non-ICP company rarely closes. This is where properly qualifying a lead does more for pipeline than any scoring tweak.
4. Look at engagement depth, not just engagement presence. This is the check nobody else writes about and the one I would keep if I could keep only one. Presence tells you a demo was opened; depth tells you which steps they explored, how far they got, what they replayed, and where they lingered. A prospect who replayed your integrations flow twice is a different animal from one who bounced after the first screen, and tracking exactly which parts of a demo a prospect explores turns a flat score into a confidence read. Wiring that depth signal back to closed revenue is how you prove it works, which is the whole point of tying demo analytics to revenue attribution.
5. Screen for disqualifying negative signals. Some signals should override a high score outright, no matter how many points sit above them. A competitor domain, a job title clearly outside your ICP, or a recent unsubscribe are all reasons to hold fire. Build these as hard negatives in your model so a genuine red flag cannot be buried under behavioral noise.

Turning a Verified Intent Score Into Action

A verified score is worth acting on precisely because verification is scarce. Salesforce research has found reps spend well under half their time actually selling, so every hour handed to a false positive is stolen from a real deal. The tiering below only works once the checks above are satisfied, which is the difference between this table and the generic cold/warm/hot chart every vendor prints.

Notice that "hot" is not a score band, it is a set of passed checks. An account only earns direct outreach once it clears corroboration, timing, ICP fit, and engagement depth with no disqualifying signals. For account-level plays this maps cleanly onto account-level intent scoring in an ABM motion, where the buying group matters more than any single contact.

The other benefit of tiering off checks rather than the raw number is that it makes your handoffs auditable. When a rep asks why an account was routed to them, the answer is a list of satisfied conditions, not a black-box digit nobody can explain.

That transparency is what rebuilds rep trust in a scoring model that has burned them before. Tie your actions to checks, not to the number, and your reps stop chasing ghosts.

Signals Worth Combining With Your Intent Score

No single data source is trustworthy enough to act on alone, which is why the smartest teams blend three tiers and weigh each by how reliable it is. First-party data comes from your own properties: site behavior, product usage, demo engagement.

Second-party data is someone else's first-party data shared directly with you, like a partner's event attendee list. Third-party data is aggregated category intent sold by publishers and networks.

The reliability tradeoff runs in the opposite direction from the volume. Third-party data is broad but noisy and often account-level guesswork, because most buyers do the bulk of their research independently before they ever talk to a vendor, so much of their intent is invisible to you and third-party providers are inferring it.

First-party data is narrower but far more trustworthy because you observed it directly. The discipline is to use third-party data to widen your net and first-party data to confirm, never the reverse, an approach I dig into further under intent-based marketing.

The most underused first-party signal is product and demo engagement analytics. Reps too often work off a bare form fill, learning only that somebody submitted an email, with no read on what that person actually did next.

"Reps can't access HubSpot. It's just a marketing tool."

[sales enablement, cybersecurity research]

That gap is exactly why depth-of-engagement data is so valuable: it survives even when the score, the CRM, and the enrichment stack all disagree, because it is a direct record of what the buyer chose to explore. One operations leader described building their qualification on precisely that signal.

"We do use it a lot for just to see the activity folks on the website are taking and engaging with demo. We use it for our lead scoring."

[operations manager, SMB business-management software]

For that team, engagement-based scoring is not a theory, it is a live part of their monthly motion, and a solid chunk of their qualified leads come from it every month. That is the bar to aim for: a signal so concrete you route real pipeline on it.

Common Mistakes That Make Intent Scores Unreliable

Most broken scoring models fail for the same handful of reasons, and every one of them is fixable. If your reps have stopped trusting the score, start here.

• Static models that never get revisited. A weighting set you defined a year ago is scoring today's buyers on yesterday's behavior, and it quietly drifts out of reality.
• No decay function. Without expiry, an old spike counts forever, so accounts that went cold months ago keep sitting at the top of the queue.
• Ignoring negative signals. A model that can only add points will always overstate interest, because it has no way to represent a red flag like a competitor domain or an unsubscribe.
• Single-signal reliance. Scoring almost entirely off one behavior, usually email opens or page views, is the fastest route to a dashboard full of false positives.
• Batch-only updates. If the score recomputes overnight, your reps are acting on a snapshot that is already stale by the time they see it, and speed-to-lead suffers.

A related question worth asking of any new signal before you adopt it: does it tell you something your current stack cannot? If a data source only re-states what your CRM and enrichment tools already know, it adds cost without adding clarity. Demand a signal that is genuinely independent, and insist on knowing how it gets captured and reconciled into your CRM so intent is not lost or double-counted when it lands.

Where Storylane Demo Suite Fits

Full disclosure: this is us. I want to explain the mechanism plainly rather than sell you a dashboard, because the mechanism is the whole point of the engagement-depth argument above.

Storylane Demo Suite captures per-viewer engagement inside your interactive demos and Demo Hubs: which steps a prospect viewed, how far they progressed, what they replayed, and how long they spent on each screen. That depth signal is exactly the corroborating evidence the framework's point four calls for, and it feeds into your CRM and scoring so a high number can be checked against what the buyer actually explored. Instead of "demo opened," you get "watched the integrations flow twice and replayed pricing," which is a confidence read a rep can act on.

Now the honest limits. Storylane only sees engagement on the demos and product experiences you have instrumented, so it says nothing about a buyer researching you in a Slack community or on a review site. It is not a third-party intent provider and it is not a firmographic data source, so it will not tell you a company's headcount or that they are in-market broadly.

It is a corroborating first-party signal, not a standalone scoring engine. The right way to use it is as the verification layer on top of the model you already run, not as a replacement for it.

FAQ

What is a good intent score threshold for sales outreach?

Most models treat the 60 to 75 band as sales-ready, but that number is a convention, not a guarantee. Treat any threshold as the point where verification starts, not where outreach starts. The right threshold for your team is the one where corroborated, ICP-fit signals reliably convert.

How is an intent score different from a lead score?

An intent score estimates whether an account is actively evaluating a purchase right now, so it is mostly about timing. A lead score measures the fit and readiness of an individual contact, so it is mostly about who to work first. Good teams use them together: intent for when, lead score for who.

How often should intent scores update?

As close to real time as your stack allows, because intent decays fast and batch-only updates leave reps acting on stale snapshots. At minimum, apply a decay function so old signals lose weight automatically. If your model only recomputes overnight, treat every morning's "hot" list as already a day behind.

Can a high intent score still be wrong?

Yes, and pretending otherwise is how reps waste time. Competitor research, analysts, bots, and early-stage procurement scans all generate high scores without any real buying intent. That is exactly why the five-point verification framework exists: to catch the false positives before a rep acts.

What's the difference between first-, second-, and third-party intent data?

First-party data is what you observe directly on your own properties, and it is the most reliable. Second-party data is another company's first-party data shared with you, such as a partner's event list. Third-party data is aggregated category intent from outside publishers, which is broad but noisy and best used to widen your net rather than to confirm a specific buyer.

Sources

• Salesforce, State of Sales (7th edition), 2026

Want to see engagement depth in action? Start for free with Storylane and watch exactly how prospects explore your demo, then feed that signal into your intent score so you always know whether a prospect is actually interested.

Prefer a walkthrough of how Storylane Demo Suite turns demo engagement into a verifiable intent signal for your team? Book a demo and we will show you the engagement-depth data live.