Intent Data Analysis: A Scoring Framework That Works Without a Data Team

Intent data analysis is the process of turning behavioral signals — website visits, content downloads, review-site activity, buying-committee movement — into a single score that tells a rep who to contact next and why. Most guides stop at defining the signals. That's the easy part. The part that actually changes pipeline is deciding how much each signal is worth, how fast that value decays, and where the line sits between "worth a look" and "call today." You don't need a data science hire to build that. You need three input categories, a spreadsheet or a CRM property, and the discipline to adjust the weights instead of rebuilding the whole model every quarter.

What Actually Counts as an Intent Signal

Intent signals split into three tiers, and they are not interchangeable:

• First-party signals — behavior on your own site and product: pricing page visits, repeat sessions, person-level identified visitors, integration or security page views.
• Second-party signals — activity on platforms where buyers compare vendors directly, like G2 category and comparison views.
• Third-party signals — category-level research aggregated across publisher networks (Bombora and similar), showing a company is researching a topic somewhere on the web, not necessarily on your site.

First-party signals carry the highest signal-to-noise ratio because they're specific to your product, not your category — see The 5% Rule for the full argument. Third-party data is still useful, but for a different job: surfacing accounts you should watch, not accounts that are ready now.

Why Counting Signals Fails You

The default mistake is treating intent scoring as addition: tally every signal event and rank accounts by volume. That rewards noisy accounts over quiet high-intent ones. A single visitor clicking through a retargeting ad ten times racks up more "events" than two named stakeholders each visiting your pricing page once — but the second account is the one with a buying committee forming, and the first is one person doing idle research.

Volume isn't intent. Distinct, ICP-fit people taking high-commitment actions is intent. A scoring model has to weight who and what page before it weights how many times.

A Scoring Framework You Can Build This Week

Score every identified account across three categories, then sum to a single composite number:

• Fit (0–30 points) — company size, industry, and tech stack match against your ICP. This is static data you likely already have in your CRM.
• Engagement (0–50 points) — the heaviest-weighted category. Give more points for distinct identified contacts engaging than for repeat visits from one person, more points for high-intent pages (pricing, integrations, security) than top-of-funnel content, and apply weekly decay so a visit from six weeks ago counts for less than one from yesterday.
• Trigger events (0–20 points) — funding rounds, leadership changes, hiring surges, or a G2 comparison view against a named competitor.

Map the composite score to an action tier:

• 🔴 Very High (80–100) — dash to rep, contact within the hour
• 🟠 High (60–79) — rep alert, contact same day
• 🟡 Medium (40–59) — watchlist, monitor for a second signal
• 🟢 Low-Medium (20–39) — marketing nurture only
• ⚪ Low (0–19) — suppress, don't burn rep time

None of this requires machine learning. It requires a property on the account record, a handful of if/then rules, and a review cadence. If you're using a platform built for GTM automation, this scoring model is the filter logic behind a workflow: identify the visitor, filter on fit and engagement thresholds, and route straight to Slack or your CRM when an account crosses into the top tier — no manual scoring spreadsheet required once it's built once.

Ground the Weights in Real Behavior, Not Just Volume

When we looked at our own closed-won pipeline, most deals moved through a single contact start to finish — normal for a lower-touch, self-serve motion. But the larger, more considered deals were noticeably more likely to involve a second or third stakeholder engaging before close than the smaller, faster ones. That's exactly why the Engagement category above weights "distinct contacts from the same account" higher than "total page views by one person": a second stakeholder showing up isn't noise, it's the buying committee assembling. For the mechanics of spotting that cluster in real time, see How to Detect a Buying Committee Forming on Your Website.

Common Mistakes That Break a Scoring Model

• Treating all signal tiers as equal. A first-party pricing-page visit and a third-party category surge are not the same strength of signal — weight them differently or your top tier fills with accounts that are merely category-aware, not product-ready.
• Skipping recency decay. A 45-day-old spike isn't a hot account anymore. Decay weekly, not per quarter.
• Scoring contacts instead of accounts. One contact visiting five times and five contacts visiting once are very different signals that a contact-only score can't distinguish — score at the account level and roll contacts up into it.
• Never revisiting the weights. Pull your last quarter's closed-won accounts and check whether your top-tier scores actually correlated with what closed. If they didn't, adjust the point values — don't rebuild the whole framework.

Build vs. Buy for the Underlying Data

The framework above is only as good as the signals feeding it:

• First-party identification — a website visitor identification platform that matches traffic against an identity graph built from a consent-based publisher network. Knock2 publishes 93%*/62%* identification benchmarks (account-level / person-level) measured against engaged sessions, so you know what denominator any "match rate" claim is built on before you trust it.
• Third-party category intent — providers like Bombora for discovering accounts researching your category before they ever hit your site.
• Orchestration — a CRM custom property or a dedicated GTM automation platform to hold the score, apply decay, and trigger the routing action once an account crosses tier.

For the case that first-party should anchor the model rather than third-party, see First-Party Intent Data: Your Highest-Converting SDR Trigger. For what happens after an account crosses into the top tier, see The BDR Playbook for Website Visitors.

FAQ

What's the difference between intent data and basic website analytics?

Website analytics tells you what happened on your site — page views, sessions, bounce rate. Intent data analysis interprets those events (plus off-site signals) to estimate how close an account is to buying, and turns that estimate into a score you can act on.

How many intent signals do you need before acting?

One signal is rarely enough to justify a sales touch on its own. The framework above solves this by scoring across three categories rather than waiting for a magic single trigger — an account can earn a Very High score from strong fit plus multiple engaged contacts even without a dramatic single event.

Do you need a data science team to build an intent scoring model?

No. A weighted point system across fit, engagement, and trigger categories, held in a CRM property or spreadsheet, covers the vast majority of use cases. Machine learning models can refine this later, but they're not a prerequisite for a working scoring model.

How often should you update signal weights?

Quarterly at minimum — pull your closed-won accounts and check whether your top-tier scores correlated with what actually closed, then adjust point values accordingly. Don't wait a year; buying behavior and your ICP both shift faster than that.

Can intent data analysis help with upsell and retention, not just new business?

Yes. The same framework applied to existing customers — a spike in engagement on a competitor's page, a new stakeholder from an existing account visiting your pricing page — is an early signal of expansion opportunity or churn risk.

Score Your Own Traffic, Starting Today

A scoring model is only useful if the underlying signals are accurate. If you're not identifying who from an ICP-fit account is actually on your site at the person level, you're scoring on incomplete data.

Start identifying your website visitors with Knock2 →

*Identification rates measured against engaged sessions. Results may vary by traffic profile, geography, and industry.