Lead Scoring Calculator
Scores a lead against your own criteria, keeps fit and engagement apart, and checks whether the tiers actually close at different rates.
Score 39.0 of 119.0 available
as a share 32.8%
tier not yet qualified
qualified at 40.0
sales ready at 60.0
Fit against engagement
fit 25.0 of 58.0, 43.1%
engagement 14.0 of 61.0, 23%
which is a lead whose fit and engagement roughly agree, which is the only case where a single score means much
Disqualifiers that applied
Free email domain -15.0
What is missing, largest first
Booked a demo 25.0 points, which would take the score to 64.0
Clicked a pricing link 12.0 points, which would take the score to 51.0
Industry we have references in 10.0 points, which would take the score to 49.0
Replied to a sequence 10.0 points, which would take the score to 49.0
Uses a competitor we replace 8.0 points, which would take the score to 47.0
Does the model separate anything
not yet qualified 1.2% close rate, 0.39× the baseline
qualified 4.8% close rate, 1.55× the baseline
sales ready 14.5% close rate, 4.68× the baseline
spread 12.08× between the top and bottom tier
verdict the tiers rise and the top closes at least twice the bottom, so the model is sorting on something real
The model
criteria 11
met 6
positive points available 119.0
negative points available -15.0
The score is 39.0 of 119.0, which puts this lead in "not yet qualified".
The arithmetic is the easy part: whether the weights predict anything is
the question, and only close rates by tier can answer it.
The close rates above are the test. A model whose tiers close at similar
rates is sorting leads without predicting anything, and the fix is
fitting the weights to outcomes rather than adjusting the thresholds
until the tiers look balanced.
Fit and engagement answer different questions and should stay apart. A
perfect-fit company that has done nothing is a marketing problem; an
enthusiastic lead who cannot buy is a distraction. One combined number
describes both as average.
The disqualifiers did real work here. Negative points are usually more
predictive than positive ones, because "not obviously wrong" is a weaker
signal than "definitely wrong": a free email domain, a competitor, or a
country you cannot serve.
Engagement decays. An email opened nine months ago is not evidence of
anything, so points for behaviour need an age limit or the score drifts
upwards forever and every old lead eventually qualifies.
Thresholds are a sales capacity decision as much as a quality one. The
right qualified line is the one that produces roughly as many leads as
the team can call, and moving it is a staffing conversation rather than
a scoring one.
A model nobody uses is worse than no model. If sales ignore the tiers,
the argument is about trust rather than arithmetic, and the fastest way
to win it is showing the close rate by tier on their own historical
deals.
Output is valid and updates as you type.
Fix the highlighted fields to update the output.
Scoring a lead is trivial arithmetic: add up the points for what is true about them. The part that decides whether the score means anything is where the weights came from.
A model built by asking the sales team what a good lead looks like is a record of their opinion. It will sort leads into tiers, and the tiers will close at roughly the same rate. The only test is your own history: score the last few hundred closed deals as they were at the time and compare the close rate of each tier. If they do not separate, the model does not work, however sensible the criteria sound.
Two structural choices are built in. Fit and engagement stay apart, because the right company doing nothing and an enthusiastic lead who cannot buy are different problems and one number describes both as average. And negative points are supported, because disqualifiers predict better than positive signals: “definitely wrong” is a stronger signal than “not obviously wrong”.
How to use
- Write the fit criteria, one a line, as a label and points. Mark the ones this lead meets with a
*. - Do the same for engagement criteria, which are behaviours rather than attributes.
- Put in your close rate by tier from real closed deals. That block is the one that says whether any of this is worth doing.
Example
A lead meeting four fit criteria including one disqualifier, and two engagement criteria:
Score 39.0 of 119.0 available
as a share 32.8%
tier not yet qualified
Fit against engagement
fit 25.0 of 58.0, 43.1%
engagement 14.0 of 61.0, 23%
Disqualifiers that applied
Free email domain -15.0
What is missing, largest first
Booked a demo 25.0 points, which would take the score to 64.0
Clicked a pricing link 12.0 points, which would take the score to 51.0
Does the model separate anything
not yet qualified 1.2% close rate, 0.39× the baseline
qualified 4.8% close rate, 1.55× the baseline
sales ready 14.5% close rate, 4.68× the baseline
spread 12.08× between the top and bottom tier
verdict the tiers rise and the top closes at least twice the bottom, so the model is sorting on something realA twelvefold spread between the top and bottom tier is a model worth trusting. A spread of 1.2 would mean the scoring is an elaborate way of sorting a list alphabetically.
Pitfalls
A model not fitted to outcomes is an opinion. The close-rate block is not optional decoration. Without it there is no way to tell a predictive model from a confident one.
Fit and engagement are different questions. Keep them apart. A perfect-fit company that has done nothing is a marketing problem; an enthusiastic student is a distraction. Combined into one score, both look average.
Most models are missing the negative half. A competitor, a free email domain, a country you cannot serve, a job title that never buys. Disqualifiers do more work than any positive signal and most models have none.
Engagement decays. An email opened nine months ago is not evidence. Without an age limit the score drifts upwards forever and every old lead eventually qualifies.
Thresholds are a capacity decision. The right qualified line produces about as many leads as the team can call. Moving it is a staffing conversation, not a scoring one.
Adjusting thresholds does not fix a bad model. If the tiers do not separate, the weights are wrong. Moving the line only changes how many leads are mislabelled.
A model nobody uses is worse than none. If sales ignore the tiers, the problem is trust, and the fastest way to fix it is showing them the close rate by tier on their own deals.
Compatibility
Arithmetic in the browser: nothing is uploaded and nothing is stored.
Criteria are read from text so the model is yours rather than a fixed template. A line marked with *,
x, +, [x] or yes counts as met, [ ] and an unmarked line do not, and the points are the last
number on the line so a label may contain digits (“Company size 50 to 500: 15”).
Negative points are kept separate in the summary, so the available positive total is not distorted by disqualifiers, and the score can legitimately be lower than zero.
The verdict on the model has three states rather than two: tiers that rise with a spread of at least two times, tiers that rise weakly, and tiers that do not rise in order. The middle case is the common one and it is the one a single pass-or-fail check would report as a success.