Lead scoring

Lead scoring is the practice of ranking prospective customers by their predicted likelihood of converting, so that limited sales attention is spent on the leads most likely to be worth it.

A score is produced by a model trained on outcomes: which attributes and behaviors preceded conversions in the seller’s own records. Typical inputs are firmographic attributes such as size, sector and source, together with engagement signals recorded against the lead.

The output is a probability, but the value is the ordering. A founder with four hours and four hundred leads does not need to know a lead’s exact conversion probability; they need to know which twelve to call today. Scoring is a sorting problem expressed as a prediction.

Two limits belong in any honest definition. A score is not qualification — it ranks, it does not judge whether a customer is worth serving. And scores are weak at cold start, because a business with no history of outcomes has given the model nothing specific to learn from.

lead-scoring is one of the deployable models in the newc0 model library. See also sales forecasting.