Lead Scoring Model

What this tool does

CM8-349 scores leads on two things — how well they fit what you sell, and how much they have actually engaged — combines them into one number, grades them A to D, and then does the part most scoring models never do: it checks the grades against your own won and lost record to see whether they predict anything at all.

Everything runs inside this single file. There is no account, no upload and no network request of any kind, so company names, contacts and deal values never leave the computer you are using.

Why score leads at all

Not to be scientific. To stop good leads going cold while somebody works through the list in the order it arrived. A scoring model earns its place if it changes what gets worked on first — and if it does not, it is administration.

That is why the calibration table exists and why it is the first table in the report. A model nobody has tested is a model nobody should be following.

The fit score

Five judgements, each from 0 to 3: right kind of business, right size, has the problem you solve, can afford it, and who you are talking to.

Fit is what the lead is. It does not move because somebody opened an email. Score it once when you know enough, and change it only when you learn something that changes the facts — a company turns out to be a tenth of the size you thought, or the person you are talking to turns out to sign.

Be hard on the authority line in particular. "Talking to a user" is a 1 even when the user is enthusiastic, and enthusiasm from somebody who cannot buy is the most common reason a promising lead dies quietly.

The engagement score

Four components, out of 100:

Count only meaningful interactions: replies, calls answered, meetings, a real reading session. Emails you sent are not engagement, and neither is an open. If you count sends, every lead you chase hardest will score highest, which inverts the whole point.

Engagement decays because it should. A lead that asked for a quote four months ago and has not spoken since is not a warm lead, however many things they did in week one.

Closed leads are an exception, and it matters. For anything won, lost or disqualified, recency is measured against the day it closed rather than today. Without that, every lead you won would quietly lose engagement points as the months passed, a deal that graded A when it was live would show as a B or C a year later, and the calibration table would be comparing grades no lead ever actually had. The grade you see on a closed lead is the grade it had at the end.

Putting the two together

The default weighting is 60% fit, 40% engagement. Weight fit higher for long considered sales where the wrong customer is expensive to serve; weight engagement higher for fast transactional sales where almost anybody could buy and the question is only whether they will.

Grades and thresholds

You set two thresholds and the grades follow:

• Grade A — at or above the higher threshold. Worth a salesperson's time now.
• Grade B — at or above the lower threshold. Worth nurturing.
• Grade C — at least half the lower threshold.
• Grade D — below that. Do not spend selling time here.

Set the thresholds by looking at leads you already know the answer to. Score ten you won and ten you lost, and put the higher threshold where the two populations separate. That takes an afternoon and is worth more than any default.

The only test that matters

The calibration table and the first chart answer one question: does grade A win more often than grade C?

Below the sample size you set, the table shows a dash and the chart leaves the grade out entirely. This is deliberate. A win rate calculated from three closed leads will read as 33% or 67% and both are noise. The default of five is a floor rather than a target — it is the point at which a rate stops being meaningless, not the point at which it becomes reliable. Twenty is where the number starts to be worth arguing about, and raising the setting as your record grows is the right thing to do.

What good looks like: a clear ladder, A above B above C above D. What bad looks like: grades that all win at about the same rate, which means the score is not carrying information, or a ladder that runs backwards, which means it is carrying information and you have the weights inverted.

Tuning the model

If the ladder is flat, change one thing at a time and watch the table:

• Move the fit weight up or down by 20 points.
• Check whether one fit component is doing nothing — if every lead scores 2 on "right industry", it is not discriminating and the real difference is somewhere else.
• Look at what your won leads have in common that is not on the list at all. That is usually the thing worth scoring, and it is usually specific to your business: a trigger event, a role change, an existing system they are stuck with.

Rescoring old leads changes the calibration, which is fine and expected. What is not fine is rescoring a lead because it was won — that guarantees a beautiful ladder that means nothing.

Fit and engagement disagreeing

The last chart shows both components for every open lead, and the interesting leads are the ones where they disagree:

• High fit, low engagement — the right customer who has not noticed you. This is where selling effort pays, and where a good lead is most often lost by waiting.
• Low fit, high engagement — interested, enthusiastic and not going to buy, or not going to be worth serving if they do. Disqualify politely and early. The sample register has one of these in it deliberately.

What a score is not

• It is not a probability. A score of 80 does not mean an 80% chance of anything.
• It is not a forecast. The pipeline total here is unweighted on purpose — multiplying deal values by scores produces a confident number built on two guesses.
• It is not a substitute for a conversation. The fastest way to score a lead accurately is to ask three questions, and no model beats that.
• It cannot see anything you have not entered. A lead with a blank last-interaction date loses twenty points for a reason that may be that nobody updated the row.

Printing and sharing

The Report tab prints the tiles, charts and both tables with a title block you fill in. Filter by source first if you are reviewing a channel, or by outcome if you are reviewing what you lost.

Saving your work

The register is held in this browser, on this computer, and stays there between visits. Use the backup button to write a JSON file you control — that file is the only copy that survives clearing browsing data or moving to a new machine. The spreadsheet download gives you the same rows to work with elsewhere.

Accuracy & disclaimer

Every weight in this tool is stated in full above, and every one of them is a choice you can disagree with. The tool applies them exactly as set and tests them against your own outcomes; it has no view about which leads are good and no data about anybody else's business.