Lead scoring
Lead scoring ranks a contact using a formula built from firmographic fit and behavioral signal, so a sales team knows who to work first instead of treating every record in the pipeline as equally worth a call.
Lead scoring ranks a contact using a formula built from firmographic fit and behavioral signal, so a sales team knows who to work first instead of treating every record in the pipeline as equally worth a call. It evaluates a person, not a company: did this specific contact visit the pricing page twice, request a demo, open three emails this week. That's a different question from account scoring, which evaluates the company itself, headcount, funding stage, tech stack, industry, and barely moves month to month by comparison.
Fit gates the model, behavior ranks inside it
Blending fit and behavior into one weighted number is the most common way a lead scoring model breaks. It lets a bad-fit account with a lot of activity outscore a great-fit account that's quiet, a fifty-person agency binge-reading a blog outranking an enterprise account where two real stakeholders just visited pricing. Running fit as a yes-or-no gate first, then scoring behavior only for accounts that clear it, keeps the two questions separate the way they need to be.
A workable model starts with five to seven attributes, not fifteen, split across firmographic, behavioral, and disqualifying categories. More attributes sound more rigorous, but when a segment converts well and still scores low, there's no way to isolate which of fifteen inputs is wrong. Decay matters too, and not at one flat rate: a demo request should hold its value across a long window, while a single blog read should fade fast, since a flat decay curve treats a strong commitment signal and a weak one as equally perishable.
In practice
Set the routing threshold from real closed-won and closed-lost data. Pull the score each account had at the moment it entered the pipeline, and find where the two outcome groups actually separate, rather than declaring a round number like 70 because it felt definitive in a meeting. Then recalibrate that threshold and the attribute weights against outcomes on a cadence, monthly for a fast sales motion, quarterly for a longer one, since a model built once and left alone quietly turns into a column nobody trusts.
What people get wrong
A score nobody can trace back to a specific, verifiable signal gets quietly ignored. Reps stop acting on a number they can't defend to a manager or a skeptical prospect, and within a few months the team is running an informal ranking in Slack while the official score sits on the CRM record as decoration. More attributes or a fancier model doesn't fix that. Fewer attributes, tied to signals a person can point to, does.
