What Each One Actually Computes
A lead score is arithmetic on an opinion. Someone decided a demo request is worth 30 points, a webinar registration 10, and a director title 15. The total is a ranking, not a probability. Two leads scoring 55 have nothing in common except that the rulebook liked them equally.
A propensity score is a probability with a target attached. It answers one question: what share of records that looked like this one went on to convert. That makes it testable. You can bucket last quarter's scores, measure actual conversion in each bucket, and see whether the model told the truth.
Why Rule-Based Scores Drift
Point values are written once and then inherited. Meanwhile the channel mix changes, a competitor shifts pricing, buying committees grow, and the content library gets rebuilt. None of that flows back into the rules.
The result is predictable. Sales stops trusting the MQL flag, marketing keeps reporting volume against a threshold that no longer separates buyers from browsers, and the handoff argument repeats every quarter. A propensity model has the same exposure to a changing market, but retraining fixes it. Rewriting a point rulebook rarely happens at all.
When Each One Is the Right Call
Use rule-based scoring when you lack resolved history, when the buying motion is brand new, or when regulators and stakeholders require a score they can read line by line. The transparency is real and it matters.
Use a propensity model when you have enough closed outcomes to learn from and enough volume that manual triage wastes seller time. The model finds combinations no analyst would have written down, such as a mid-tier title that converts far above average when paired with a specific product page visit.
Running Both Without Conflict
Keep the rulebook as a floor and the model as the ranking. Hard rules handle disqualification, which is a policy question rather than a prediction: wrong geography, competitor domain, existing customer, unsupported deployment. The propensity score then orders everything that survives.
Report both against the same denominator so the comparison stays honest. Track conversion to qualified opportunity and eventual win rate by score band for each method over the same period. Whichever method separates the bands more cleanly should own routing. Feed the winner into the same pipeline math you use in how to forecast revenue, because a score that never reaches the forecast is a reporting artifact.
Frequently Asked Questions
What is the core difference between lead scoring and a propensity model?
Lead scoring uses weights a human assigns, such as 10 points for a demo request. A propensity model derives its weights from historical conversion outcomes, so the weights reflect what actually predicted revenue instead of what someone assumed would.
Can a propensity model replace lead scoring entirely?
It can once you have enough resolved records for the model to learn from. Before that, rule-based scoring is the reasonable fallback because it needs no training data and it forces the team to state its assumptions explicitly.
Why do rule-based lead scores drift?
Point values get set once and almost never revisited, while channel mix, buyer behavior, and content library all change underneath them. The score keeps rewarding actions that stopped correlating with revenue, and sales quietly stops trusting the MQL flag.
What outcome should a propensity model predict?
Pick one outcome and name it precisely, such as becoming a qualified opportunity within 60 days or closing won within two quarters. A model trained against a vague target like sales-ready produces a score nobody can act on.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like propensity model vs lead scoring into prescriptive action for your team.
Schedule a Demo