Lead scoring: a free calculator and how to set the points

This page is for marketing ops, sales ops and founders who have more leads than their reps can work properly and need a fair way to decide who gets a call today. The calculator above scores one lead at a time. The rest of the page shows how to replace its illustrative points with numbers from your own pipeline, test the thresholds on last quarter's leads and keep the model honest once it's live. You'll need a CRM export of last year's leads and about two hours.

Fit, intent and engagement measure different things

Lead scoring adds points for evidence that a lead will buy and subtracts them for evidence that it won't. The evidence comes in three kinds that age at different speeds:

Keep fit visible even inside one total. A perfect-fit company with no engagement needs outreach, not a nurture email. An eager contact at a poor-fit company needs one qualifying question, not a sales cycle. The big CRMs keep the two apart: HubSpot's lead scoring tool can label a combined score with a fit letter (A to C) and an engagement number (1 to 3), and Salesforce Account Engagement gives each prospect a letter grade for fit alongside a score for activity.

The fit and intent rows are the hardest to fill honestly, because the answers sit in company news, hiring pages and filings, not your CRM. Bebop researches each account and reports what's happening there and who to talk to, so those rows can be ticked from evidence; check its facts as you'd check your own. For which signals deserve intent points, see what makes a B2B lead high intent.

How the calculator works

Tick what's true for one lead. Each ticked row adds its points to the score at the top, and every number, including the negatives and both thresholds, is editable. The default points are illustrative, not a calibrated prediction: a score isn't a chance of buying until you've set the points from your own data.

• Bands. A score at or above the sales-ready threshold (default 70) reads "Sales-ready: route to a rep". At or above the nurture threshold (default 40) it reads "Nurture: worth a researched touch". Anything lower reads "Not now: keep out of the sales queue".
• Copy this model copies every row with its tick and points, the thresholds, and this lead's score and band as plain text for a doc or CRM note.
• Reset to the example restores the default points, ticks and thresholds. Nothing is saved, so reloading the page resets it too.

Two properties of the defaults matter. Fit maxes out at 60, so no lead reaches 70 on fit alone. But intent and engagement together can reach 80, so a lead with no fit at all can score as sales-ready. To prevent that, add a fit floor to your routing rules, such as 30 fit points.

Worked example: three leads scored with the defaults

Illustrative example. The companies and people are fictional.

Tallyframe (tallyframe.example) sells accounting software to companies that run many legal entities: property groups, franchise groups and holding companies in the US and Canada with 50 to 1,000 employees. Three leads arrive in the same week.

Lead A is the calculator's starting example. Its software is unknown, so the tool row stays unticked: score unknowns as zero, don't guess.

What makes this work:

• Lead A gets a call today, and the acquisition is the reason to call.
• Lead B is one signal away: a reply or a seventh location (+15 either way) takes her to 75. Nurture here means one researched email about running six sets of books, and a question about who signs off on software.
• Lead C shows why negative scores exist. Without the -50, the analyst scores 50 and lands in nurture, one click from a rep's queue.

Negative scores: rules first, points second

Disqualifiers (competitors, students, existing customers, open deals) don't make a lead less likely to buy. They put it in the wrong queue. Handle them as routing rules before any scoring: exclude competitors and students, and send customers to their account manager. The calculator gives them large negatives, but points can't guarantee exclusion: a competitor who ticks every positive row still scores 90.

Soft negatives (a personal email address, "no budget this year") lower the odds without ruling a lead out. Size them from data, the same way as positive points in the next section. If personal-email leads convert at a quarter of your average rate, the evidence says -20, not -10.

Score decay: let old signals fade

Intent and engagement points should shrink as they age. Fit points shouldn't, unless the fact behind them changes. A simple rule: halve engagement points after every 90 days with no new activity, and untick a trigger once it's more than a quarter or two old.

HubSpot's lead scoring tool has a decay setting for each group of scored events, and its example reduces scores by 50% every three months. Salesforce's Einstein Behavior Scoring has decay built in. The calculator doesn't decay anything, so apply it by editing points.

Say Lead A's controller goes quiet. Four months later the acquisition is old news, so the trigger comes off (-15), and the reply has halved once, from 15 to 8 (-7). The score is 80 - 15 - 7 = 58: nurture. To see it in the calculator, untick the trigger and set the reply's points to 8.

Set point values from your own data

Replace the default points with lift: how much more often leads with an attribute converted than leads in general.

1. Pull leads with known outcomes. Use leads old enough to have finished a sales cycle (for a 90-day cycle, leads from 3 to 15 months ago). Count one outcome: a sales-accepted opportunity, or a customer if you have well over a hundred wins.
2. Work out the baseline. Leads that converted ÷ all leads.
3. Work out each attribute's rate. Of the leads that had the attribute when they arrived, the share that converted.
4. Calculate lift. The attribute's rate ÷ the baseline. Above 1 is a good sign, below 1 a bad one.
5. Turn lift into points. Give 10 points for every doubling: points = 10 × log2(lift). In Excel or Google Sheets, that's =ROUND(10*LOG(lift,2),0).

The table is also a sanity check: 25 points for a demo request assumes demo requesters convert at nearly six times your average. If yours don't, 25 is too many.

Why doublings rather than plain multiples? Points add, but lifts multiply. If controllers convert at 2.8 times the baseline and companies in your industries at 2.2 times, a controller in one of those industries is about 6 times as likely, not 5 (2.8 × 2.2 = 6.2), if the two are unrelated. Doubling-based points add the same way: 15 + 11 = 26, the points for a lift of 6.2.

Illustrative example, continued. Tallyframe exports the 2,400 leads it created from January to June 2026. 120 became opportunities within 90 days, a baseline of 5%.

What Tallyframe learns: webinar attendance is noise, so it gets nothing. Almost every lead is in region, so region can't separate leads, but the 120 out-of-region leads produced no opportunities, so that becomes an exclusion rule. Personal email is a stronger warning than the default says, and size is weaker.

Three checks before you trust a number:

• Sample size. If fewer than about 10 leads with the attribute converted, the rate is a rough guess. Keep its points modest and look again next quarter. (Personal email has only 4 conversions, but even double that would be far below the baseline.)
• Overlap. Attributes that travel together get counted twice. Compare the rate for leads with both against leads with only one, and if having both adds little, cut one row's points.
• Hindsight. Score each lead as it looked when it arrived. Fields filled in later leak the outcome into the model; Salesforce's own setup guide suggests excluding fields such as the reason a lead didn't convert.

Period: leads created [start date] to [end date] (old enough to have reached an outcome)
Outcome counted: [sales-accepted opportunity / customer] within [90] days
Total leads (A): [   ]   Converted (B): [   ]   Baseline = B ÷ A = [   %]

One row per attribute, as known when the lead arrived:
Attribute | Leads with it (C) | Converted (D) | Rate = D ÷ C | Lift = rate ÷ baseline | Points = ROUND(10*LOG(lift,2),0) | 10+ conversions? (y/n)

Rules applied before scoring (no points):
- Competitor or student: exclude
- Existing customer or open deal: route to [account owner]
- Outside [the regions you serve]: exclude

Set the thresholds by back-testing last quarter

1. Take last quarter's leads. Use the latest full quarter whose leads have had time to convert.
2. Score each one as it looked on arrival, using your new points.
3. Group them into score bands and count the leads and conversions in each.
4. Write down rep capacity: how many new leads your reps can follow up properly in a quarter.
5. Set the sales-ready threshold at the band boundary where the leads above it fit that capacity and still convert well above the baseline.
6. Set the nurture threshold where conversion falls close to zero.

Illustrative example, continued. Tallyframe scores the 600 leads it created from April to June 2026, which produced 30 opportunities.

Tallyframe's one inbound SDR can give about eight new leads a week a proper follow-up, roughly 100 a quarter. A threshold of 70 sends 90 leads that produced 19 opportunities: 21% conversion and 63% of all opportunities. Dropping to 60 sends 160 leads, far over capacity, and the extra 70 convert at 6%, barely above the 5% baseline. Raising it to 80 leaves capacity idle and misses 7 opportunities. Sales-ready is 70.

For nurture, the 230 leads scoring 40 to 69 produced 9 opportunities (3.9%), enough to justify a researched email. The 280 below 40 produced 2 (0.7%), so they get automated content only. Nurture is 40.

Then check for leaks: 11 of the 30 opportunities came from leads below 70. Read those 11. If several share something the model underrates, such as the same referral partner, that's your next points change.

Model version: [v1, date]   Owner: [name]
Order: apply exclusions first, then hand-raises, then the score.

1. Exclude: competitors, students, outside [regions]. Existing customers and open deals go to [account owner].
2. Hand-raises go to sales whatever the score: demo request, contact-sales or pricing form.
3. Sales-ready ([70]+) with at least [30] fit points:
   Route to [SDR queue]. First touch within [1 business day].
   A rep can reject with a reason: wrong fit / bad data / already in touch / other.
4. Nurture ([40] to [69]): owned by [marketing]. One researched email, then [sequence].
   Moves to sales-ready when the score reaches [70] or on any hand-raise.
5. Not now (below [40]): automated content only. Re-score monthly.

When not to use lead scoring

For scale, Salesforce won't build an Einstein model from your data alone until 120 leads from the last 200 days have converted. A hand-built model resting on fewer is still useful, but treat it as rough.

Predictive lead scoring in HubSpot and Salesforce

Predictive scoring replaces your points with a model trained on your CRM history. It can find patterns you'd miss, at the cost of control, and each vendor sets data and plan requirements.

Even with plenty of history, keep exclusions and routing as rules you can explain. When your points and the prediction disagree sharply about a lead, read the lead.

Review cadence

Common mistakes

• Scoring engagement only. Email opens and page views reward curiosity. Fit should carry real weight.
• A threshold set to hit a lead target. Lowering the bar to send sales more leads fills the queue with leads that convert near the baseline.
• A model marketing owns alone. Sales sees the leads that fail. Their rejection reasons are your best data.

Self-check

• Every point value comes from your own lift data, or is marked as a guess.
• Disqualifiers are routing rules, not only negative points.
• On last quarter's leads, the sales-ready threshold sends a volume your reps can handle.
• Engagement points decay.
• A rep can say in one sentence why a lead is sales-ready.
• The next review date is in a calendar.

Next step: the fit rows are only as good as your definition of fit. Write it down with the ideal customer profile template, then prepare for your sales-ready leads with the sales call planning template.