Why Your Intent Data Isn't Closing Deals (And How to Fix It)

Intent data was supposed to end cold outreach. Buy a subscription, identify accounts 'surging' on your category, point your SDRs at them, watch pipeline appear. Most B2B teams have lived the uglier version: a firehose of signals that floods sales queues, burns rep credibility, and produces conversion rates barely distinguishable from a cold dial list. The data isn't the problem. The absence of a scoring layer that separates real buying behavior from casual browsing is the problem. Here's how the teams getting results are building that layer.

The Noise Problem Nobody Talks About

Third-party intent providers flag an account as surging when its employees consume more content than usual on a topic cluster over 30 or 60 days. Two structural flaws break this in practice. Keyword clusters are broad by design, so one label captures the Fortune 500 mid-rip-and-replace alongside the solopreneur reading beginner blogs. And the surge threshold compares each account to its own history, meaning a company that rarely reads anything looks active the moment one employee clicks one article.

The result is predictable. You get a 10,000-account list, 3,000 show up as active, reps work all of them, and meeting conversion hovers around 0.8 percent. That's the same rate you'd get from a cold dial with no data at all. The fix isn't a different vendor. It's a filtering layer that surfaces accounts where multiple independent signals line up in the same short window.

Signal Convergence: The Only Number That Matters

One intent spike is noise. A topic surge plus a G2 category visit plus two relevant job postings plus a recent funding round is a pattern. When three or more independent signals land on the same account inside a 14-day window, meeting conversion jumps to 4 to 6 percent. That's a five to seven times lift, and it's the entire game.

• Tier 1 (highest value): pricing page visits, trial activations, in-product usage
• Tier 2: G2 or Capterra category views, competitor comparison pages
• Tier 3: webinar attendance, gated content downloads with form fills
• Tier 4: Bombora or TechTarget keyword surges
• Tier 5: job postings, LinkedIn engagement, funding news

Build the Model From Your Own Wins

Vendor scores are black boxes. Building on top of them just stacks another black box. Instead, pull 12 months of closed-won accounts from your CRM and retroactively identify which signals were present at opportunity creation. The signals that show up in wins and not in losses are your high-weight variables. Start with a 100-point scale. Pricing page or trial activation gets 40 to 50 points. Content downloads around 15. G2 activity at 15. Third-party surge at 10. You need at least 150 closed-won accounts with signal history to trust your weights. Below that, treat the model as a v0 and refine quarterly.

Five Tiers, Five Different Plays

A scoring model only earns its keep when scores translate into specific actions. Five tiers cover the range from imminent purchase to dormant ICP fit. Tier 1 (80 to 100) gets same-day SDR outreach with genuine personalization. Tier 2 (60 to 79) gets a 48-hour SLA and a lighter content-led sequence. Tier 3 (40 to 59) enters marketing-qualified-account nurture. Tier 4 (20 to 39) gets low-cost awareness touches. Tier 5 is monitored but not worked.

The operational rule that holds everything together: hard-code tier assignments in your CRM and enforce the routing. Without that, reps default to working whatever looks interesting and the scoring model becomes decoration.

The Stack That Makes It Run

Four layers get you from raw signals to live sequences. Bombora or a comparable provider for third-party topic data. Your own site and product events as the first-party layer (the most predictive, the most neglected). Clay as the enrichment and scoring engine, pulling everything into a single table, computing composite scores, and writing back to HubSpot every 24 hours. HubSpot workflows handle routing: if tier equals 1, create a task, enroll the sequence, notify the manager.

• Fire a HubSpot custom event on every pricing page visit and trigger an immediate re-score
• Track demo button clicks separately from form submissions to catch abandoned intent
• Apply 50 percent decay to signals older than 21 days; drop signals past 60 days entirely
• Never escalate a third-party signal alone; require corroboration from a higher tier

Measure Precision and Recall, Not Volume

Most teams measure intent programs by activity: accounts identified, sequences sent, meetings booked. Activity isn't accuracy. Precision tells you what percentage of your Tier 1 and 2 accounts actually convert to opportunity within 90 days. Target 35 percent or higher. Recall tells you what percentage of new opportunities were flagged as high-intent before they opened. Target 60 percent or higher. Below 25 percent precision, your model isn't earning its spot over a cold list. Review quarterly, retrain with the latest closed-won data, and treat the system as a living thing.

Six Weeks From Audit to Production

Week 1: audit closed-won history and define the ICP gate. Week 2: configure your intent provider and instrument priority first-party events. Week 3: build the Clay table and scoring formula. Week 4: build HubSpot properties, tier lists, and routing workflows. Week 5: pilot with two SDRs and calibrate thresholds. Week 6: full rollout with baseline metrics locked in. Skipping the pilot is the fastest way to poison team trust in the model. Don't skip the pilot.

Intent data doesn't fail because the signals are wrong. It fails because nobody builds the layer that tells you which ones to trust.

← Back to all articles