Lead scoring works when it predicts which leads become revenue and sales believes it. Most scoring models fail the second test because they were built on guesses and never checked against outcomes. Here’s how to build one in HubSpot from a blank slate, step by step, with the trap in each step.
Step 1: Separate fit from engagement
Score two things independently: how well a lead matches your ideal customer (fit), and how actively they’re engaging (interest). A high-fit lead who’s gone quiet needs different handling from a low-fit lead frantically clicking. Blending them into one number hides that.
The trap: one combined score. It obscures the exact distinction reps need to prioritise well.
Step 2: Derive the signals from won deals, not intuition
Look at who actually bought. What firmographics did they share? What behaviours preceded the deal? Build your scoring from that evidence, not from a workshop where everyone guesses which actions “feel” important.
The trap: scoring the behaviours you wish mattered. Pricing-page visits often predict; ebook downloads often don’t — but teams weight them the other way round out of habit.
Step 3: Weight, don’t just add
Not all signals are equal. A demo request should dwarf a blog visit. Assign weights that reflect real predictive strength, and be ruthless — a long tail of tiny point values just adds noise.
The trap: death by a thousand five-point actions. The model ends up scoring activity, not intent.
Step 4: Set the threshold with sales in the room
The score at which a lead becomes sales-ready is a shared decision. Set it too low and you flood reps with junk; too high and you starve them. Agree it together, using real examples on both sides of the line.
The trap: setting the MQL threshold to hit a volume target rather than a quality one. That’s how sales learns to ignore the score.
Step 5: Validate against outcomes, then keep validating
After launch, check whether high-scoring leads actually convert better than low-scoring ones. If they don’t, the model is decoration. Review it quarterly against fresh outcomes and adjust.
The trap: build once, never check. A scoring model that isn’t validated is just opinion with a number attached.
The through-line
Good scoring is evidence-based, weighted, jointly owned, and continuously checked. Get those right and sales trusts it. Skip any one and you’ve built a number nobody uses.
Related reading
See three lead-scoring models sales quietly ignore and what a sales-qualified lead is.
Want your current model pressure-tested? Book a health check.