What is lead scoring?
Lead scoring is giving each lead a number, usually out of 100, so you can see at a glance which ones deserve attention first. It is a way of applying your lead qualification criteria consistently across many leads, rather than relying on who shouted loudest in the Monday meeting. The glossary has a one-line definition.
There are two broad ways to produce the number. You can write down the criteria and points yourself (a rubric), or you can train a statistical model on your past results (predictive scoring).
What is the difference between rubric and predictive lead scoring?
A rubric score is a set of rules you design; a predictive score is a model that learns rules from past data. The difference decides what you need before you start and what you can explain afterwards.
| Rubric scoring | Predictive scoring | |
|---|---|---|
| How it works | Points for criteria you choose | A model trained on past leads and their outcomes |
| What you need | A clear view of what a good lead looks like | Enough closed deals, won and lost, with consistent data |
| Explaining a score | Each point has a stated reason | Often hard to explain a single result |
| Changing it | Edit a criterion or a weight | Retrain the model, then check it again |
| Main risk | Your assumptions are wrong | Too little data, so the model learns noise |
Predictive lead scoring earns its keep in businesses with high lead volume and plenty of closed outcomes, such as software companies with thousands of trial sign-ups. A 25-person agency that wins a few dozen projects a year does not have that history. A model trained on it will find patterns, but many of them will be coincidences.
Why does a rubric suit a service firm better?
A rubric suits most small service firms because it works from day one and anyone can read why a lead scored as it did. The people who qualify leads already know what a good one looks like. A rubric writes that knowledge down so it is applied the same way every time.
It also makes disagreement useful. When a salesperson thinks a lead scored too low, they can point at the specific criterion they disagree with. That conversation improves the rubric. With an unexplained number, the only options are to trust it or ignore it.
How do you design a points rubric?
Design a rubric by choosing a few criteria that genuinely separate good leads from poor ones, giving each a maximum number of points, and writing down what earns full, partial and zero points. Five or six criteria is usually enough. More than that and people stop reading the reasons.
Start from your qualification checklist, then decide how much each criterion matters. The weights below are illustrative, for a B2B service firm. Yours should reflect your own work.
| Criterion | Max points | Full points when… | Zero points when… |
|---|---|---|---|
| Fit with what you do best | 25 | The request is squarely your core work, for a sector you know | It is work you do not do, or do not want |
| A real problem with a trigger | 25 | A specific, costly problem and a dated reason to act now | Vague interest with nothing prompting it |
| A route to the decision | 20 | You are talking to, or have been introduced to, whoever signs | No sign of who decides or how |
| Signs they can afford it | 20 | A realistic range is stated, or the scale of the organisation makes it likely | Expectations are far below what the work costs |
| A reachable, real person | 10 | A named contact with a work email and a clear role | Anonymous or incomplete details |
| Total | 100 |
A few rules make a rubric hold up:
- Define partial points. Write what earns, say, 10 of 20 for affordability. Without it, two people will score the same lead differently.
- Keep criteria separate. If “urgency” and “need” always move together, merge them. Double counting inflates some leads.
- Allow “unknown”. A lead with an unknown budget should score lower than one with a confirmed budget, but it should not score zero. Zero is for evidence of a poor fit.
- Set bands, not a single line. For example, 70 and above gets a reply the same day, 40 to 69 gets a reply this week, below 40 gets a polite, brief answer.
Why should a score show its reasons?
A score should show its reasons because a number on its own cannot be checked, challenged or improved. “62” tells you nothing. “Affordability 10/20: no range given, but a similar-sized organisation. Fit 25/25: matches our core service” tells you what to ask on the call.
Reasons also make the score honest about what it does not know. If most of the points were lost to missing information, the lead may be better than the number suggests, and the next step is to find out. That is a different action from a lead that scored low because it is a poor fit.
How does a rubric score work on a real lead?
Illustrative example. Take a 25-person marketing agency in Manchester. An enquiry arrives from the marketing manager at a regional chain of 12 dental practices, asking for help with a website rebuild and local search.
Using the rubric above:
- Fit with what you do best: 25/25. Website and local search is the agency’s core work, and it knows the healthcare sector.
- A real problem with a trigger: 15/25. The current site is visibly outdated. No stated deadline, but a new practice opens next spring.
- A route to the decision: 10/20. The marketing manager is likely to lead, but the owners will probably sign.
- Signs they can afford it: 12/20. No range given. The chain is a credible size, so means are likely, but unconfirmed.
- A reachable, real person: 10/10. Named person, work email, clear role.
Total: 72. It sits in the “reply the same day” band, and the reasons tell the salesperson exactly what to ask: a budget range, the opening date, and whether the owners will join the next call.
How do you recalibrate a lead scoring rubric?
Recalibrate a rubric by comparing scores with what actually happened, on a regular schedule, and changing one thing at a time. Once a quarter is enough for most small firms.
- Pull the last quarter’s leads with their scores and outcomes: pursued or not, proposal or not, won or lost.
- Look for surprises. High-scoring leads that went nowhere, and low-scoring leads that became good work.
- Read the reasons on those surprises. Is one criterion consistently wrong? Perhaps budget is scored too generously, or a sector you ranked as poor fit keeps producing work.
- Change one weight or definition, write down why, and date the change.
- Keep old scores as they were. Rescoring history hides whether the change helped. Note which rubric version produced each score.
With a small number of deals, treat the review as a conversation informed by data, not a statistical exercise. The value is in catching criteria that are plainly wrong, not in fine-tuning a point here and there.
Where OpsUPLOOP fits, and where it doesn’t
OpsUPLOOP gives incoming leads a triage score out of 100 against a rubric your firm sets: your factors, how much each one counts and what good looks like. The reason for each factor is shown alongside the research that supports it, so your team can read the reasons, disagree and act. Need help defining the rubric itself? See our sales training partner. See OpsUPLOOP Sales for how it works in your own instance.
It is not a predictive model, and its scores are not win probabilities. It will not tell you which leads will close. It applies your judgement consistently and shows its working. If you have thousands of leads a month and a data team, a trained predictive model may suit you better.
Other questions
Is a lead score the same as a win probability?
No. A rubric score ranks leads against your criteria. A win probability is a claim about the chance of closing, and a rubric score should not be read as one unless you have checked it against real outcomes.
Should the score change automatically as a lead progresses?
The initial score is best kept as a record of what you knew at first contact. Later changes in a deal are better tracked through pipeline stages than by quietly rewriting the original score.
Can AI do rubric scoring?
Yes. An AI model can read an enquiry and research and assign points against your rubric, as long as it shows its reasons so a person can check them. It is still a rubric, not a predictive model, because the criteria and weights are yours.
What score should count as qualified?
Set a threshold after scoring a batch of recent leads and seeing where the ones you would have pursued fall. Treat it as a guide for attention, not an automatic cut-off.