Lead scoring assigns a value to each lead so a team can work the most promising ones first. Useful models separate fit, meaning how closely the company matches your ideal customer profile, from engagement, meaning what that company has actually done. A score is a work order, not a prediction of revenue.
What it is
Scoring converts a pile of leads into a queue. Every rule you add is a claim about what predicts a deal, made explicit enough to argue with. Two families of signal go in. Fit signals are structural and stable: category, market, size, the technology on the site. Engagement signals are behavioral and perishable: a pricing page visit, a reply, a demo request. Blending them into a single number hides the difference between a great account that has ignored you and a poor account that clicks everything.
How it works
Start with points on the traits your best accounts shared before they bought, and subtract points for disqualifiers such as the wrong country or a category you cannot serve. Negative scoring matters more than teams expect: it is what keeps the queue honest. Then add decay, so a visit from six weeks ago stops outranking a reply from yesterday. Review the model against outcomes each month. If high scoring leads do not close at a higher rate than low scoring ones, the model is decoration.
Why it matters for winning clients
Attention is the scarce resource in outbound, not data. A team of two can research five companies properly or skim fifty badly, and scoring decides which five. It also settles arguments between sales and marketing with a rule instead of a preference. For agencies juggling delivery and prospecting in the same week, a ranked queue means the twenty minutes you have go to the account most likely to answer.
Example in LeadCanvas
A marketing agency searching independent retailers gives points for a listing with recent reviews, points for a category it has case studies in, and subtracts points for chains outside its service area. It ranks the returned companies with that rule and works the top of the list first. Because the criteria are written down, the same rule runs again in the next city and the results are comparable rather than anecdotal.
Common mistakes
Scoring on data you cannot verify, which produces confident rankings built on guesses. Treating a score as intent: a high number means worth a look, never ready to buy. Building thirty rules nobody can explain, so nobody trusts the output. And leaving the model frozen for a year while the offer, the market and the channels all move underneath it.
Related terms
The ideal customer profile supplies the fit half of the model. Intent data feeds the engagement half. Data enrichment fills the fields the rules read, and data decay is why those fields need rechecking. The links below open the tools where a scored list becomes actual work.