Lead scoring is a method of ranking leads by how likely they are to buy, so sales works the best ones first. A good model combines three dimensions: fit (does the company and person match your ideal customer profile), intent (have they taken high-intent actions) and engagement (how actively they interact), with negative points and decay keeping scores honest.
This guide explains each dimension, compares point-based and predictive scoring, and shows how to build a model from your own closed-won data with a worked, clearly hypothetical example. It also covers MQL and SQL thresholds, account-level scoring, validation, how HubSpot and Zoho CRM implement scoring today, and the mistakes that quietly ruin most models.
Key takeaways
- Score fit and interest separately. A perfect-fit company that has done nothing and an engaged student who will never buy both need different treatment, and a single blended number hides that.
- Weight high-intent actions such as demo or pricing requests far above passive engagement such as email opens, which privacy features make unreliable.
- Build points from your own closed-won and closed-lost history, not from a template. A model is a hypothesis until conversion by score band proves it.
- Use negative scoring and decay so old activity and poor-fit signals pull scores down over time.
- Agree thresholds, service levels and a review rhythm with sales. Scoring that sales does not trust gets ignored, however clever it is.
What does recent research say about buyers and lead scoring?
Buyers do more research before speaking to anyone, so the signals they leave matter more. In a Gartner sales survey released in March 2026 of 646 B2B buyers (fielded August–September 2025), 67% said they prefer a rep-free experience and 45% reported using AI during a recent purchase.
Yet buyers still want people at the right moment. A Gartner release from May 2026, drawing on 645 B2B buyers surveyed in the same period, found that 69% prefer to validate AI-generated insights with sales reps, 70% prefer a fully digital, self-service buying experience and buyers consulted an average of seven information sources for a recent purchase. For scoring, the lesson is that most of the journey is invisible, so the few strong signals you do see deserve careful weighting.
How B2B buyers research and buy (Gartner, 2026 releases)
We found no recent, independent benchmark for conversion by lead score, so this guide does not quote one; conversion depends on your market, deal size and definitions.
What is lead scoring and why does it matter?
Lead scoring assigns points based on who a lead is and what they do, then uses the total to decide the next step: nurture, sales outreach or disqualification. It sits inside the wider process of lead generation and qualification, and it only works if the hand-off rules that follow it are clear, which is the subject of our guide to lead management and sales hand-offs.
The value is focus. Without a shared ranking, reps cherry-pick by gut feel; a model makes prioritisation explicit, measurable and improvable.
What are the fit, intent and engagement dimensions?
Fit: does this lead match your ideal customer profile?
Fit uses firmographic and demographic data: industry, company size, geography, revenue band, technology used, job function and seniority. Fit should come straight from a documented ideal customer profile, and it changes slowly, so it does not need to decay.
Intent: are they showing buying behaviour?
Intent covers first-party buying actions: demo or quote requests, repeat pricing or comparison page visits, trials and replies to sales. These carry the heaviest weights. Third-party intent data (topic research across other sites, or reverse-IP company identification) adds context but is usually account-level, uneven in smaller markets and probabilistic. Use it to prioritise accounts for research, not to trigger an immediate sales call.
Engagement: how actively are they interacting?
Engagement measures the volume and recency of email clicks, downloads, webinar attendance and return visits. It shows interest and timing but is easily inflated by bots, students and competitors, so cap it and let it decay.
Should you use point-based or predictive scoring?
In point-based scoring, people choose the criteria and points. In predictive or AI scoring, software learns patterns from historical outcomes. The right choice depends on data volume and how much explanation sales needs.
| Factor | Point-based scoring | Predictive / AI scoring |
|---|---|---|
| Data needed | Works with little history; can start from ICP judgement | Needs a meaningful volume of past conversions and non-conversions |
| Transparency | Every point is visible and explainable to sales | Explanations vary by tool; usually lists top factors |
| Maintenance | Manual reviews needed as the market changes | Retrains on new data, but can drift if inputs are dirty |
| Best for | Early-stage teams, new markets, low lead volume | Established teams with clean, consistent CRM history |
Most SMBs should start rules-based, prove it with conversion data, then test a predictive score alongside it.
How should negative scoring and decay work?
Negative scoring subtracts points for signals that cut the chance of a sale: free email domains in B2B, student titles, competitor domains, unsubscribes, hard bounces and careers-page visits. Decay reduces engagement points as activity ages, so last year’s downloads do not outrank this week’s pricing request.
Fit can go negative for hard disqualifiers, while engagement should usually fall towards zero. Set a floor and ceiling per dimension so no single behaviour dominates.
How do you set MQL and SQL thresholds?
A marketing qualified lead (MQL) meets the agreed score threshold and is passed to sales. A sales qualified lead (SQL) has been accepted by sales after a conversation confirms need, authority and timing. The score drives the first status; a human decision drives the second.
Set the MQL threshold where conversion data shows a clear step up and the volume matches sales capacity. The bands below are a starting template, not a standard.
| Score band (out of 100) | Typical profile | Action | Owner and timing |
|---|---|---|---|
| 75–100 | Strong fit plus a high-intent action | Route to sales as MQL; personal follow-up | Assigned rep, same working day |
| 50–74 | Good fit with growing engagement, or high intent with partial fit | Route to SDR for qualification | SDR, within two working days |
| 25–49 | Partial fit or early-stage interest | Targeted nurture by role and topic | Marketing, ongoing |
| 0–24 | Poor fit or no recent activity | General newsletter or no action | Marketing, quarterly review |
| Below 0 | Disqualifier present (competitor, student, unsupported region) | Suppress from sales routing | Marketing operations |
How does account-level scoring support ABM?
In B2B deals several people decide together, so one contact’s score can mislead. Account-level scoring rolls up company fit and engagement across all known contacts, often adding coverage of key roles such as economic buyer and technical evaluator. This is the scoring layer behind account-based marketing, where target-account selection and prioritisation matter more than individual lead volume.
Summing contact scores rewards accounts with many low-value contacts; averaging penalises a lone active champion. Many teams use the maximum contact score plus a bonus for each engaged buying role.
How do you build a scoring model from closed-won data?
Start with outcomes, not points. Export leads covering complete sales cycles, mark which became customers, and compare each attribute’s conversion rate with the overall rate. The ratio is the lift. Attributes with high lift earn points; attributes with lift below one earn none or lose points.
Worked example (hypothetical)
Imagine a B2B software company in India with 2,000 leads in 12 months, 60 of which became customers: a 3% overall rate. All numbers are invented for illustration.
| Signal | Dimension | Leads with signal | Became customers | Conversion | Lift vs 3% | Points |
|---|---|---|---|---|---|---|
| Company size 50–500 employees | Fit | 600 | 36 | 6.0% | 2.0x | +15 |
| Target industry (manufacturing, logistics) | Fit | 500 | 30 | 6.0% | 2.0x | +15 |
| Job title director or above | Fit | 400 | 20 | 5.0% | 1.7x | +10 |
| Demo or pricing request | Intent | 150 | 33 | 22.0% | 7.3x | +30 |
| Viewed pricing page twice in 30 days | Intent | 300 | 24 | 8.0% | 2.7x | +15 |
| Clicked two or more emails in 30 days | Engagement | 500 | 20 | 4.0% | 1.3x | +5 |
| Opened emails (opens only) | Engagement | 1,200 | 38 | 3.2% | 1.1x | 0 |
| Free email domain | Fit | 700 | 7 | 1.0% | 0.3x | -15 |
| No activity for 90 days | Decay | – | – | – | – | -50% of engagement points |
Points roughly follow lift, with caps of 40 for fit, 45 for intent and 15 for engagement so the total stays within 100. Opens earn nothing.
Now score one hypothetical lead: a director at a 120-person logistics firm, using a company email, who clicked two emails and viewed pricing twice this month. Fit is 15 + 15 + 10 = 40. Intent is 15 for the pricing views. Engagement is 5. The total is 40 + 15 + 5 = 60, which lands in the 50–74 band and goes to an SDR. If the same lead then requests a demo, intent becomes 15 + 30 = 45 (at the cap) and the total rises to 90, triggering same-day sales follow-up. If instead they go quiet for 90 days, decay halves engagement to about 3 and the total drifts to 58.
How do you calibrate and validate a lead scoring model?
Validation checks that higher scores really convert better. Back-test the model on a past cohort, then track each new cohort by band through SQL, opportunity and customer. Conversion should rise steadily from band to band. If middle bands convert as well as the top band, move the threshold.
Lead-to-customer conversion by score band (hypothetical, illustrative only)
Review the model each quarter, and after any change to your offer, pricing or target market. Record the model version in the CRM; our guide to marketing analytics and attribution covers how to keep reporting definitions consistent.
How do HubSpot and Zoho CRM implement lead scoring?
As of November 2026, HubSpot’s lead scoring tool (HubSpot Knowledge Base, updated August 2026) sits under Marketing > Lead Scoring and is available with Marketing Hub or Sales Hub at Professional or Enterprise tier; contact scores require Marketing Hub and deal scores require Sales Hub. It supports fit scores built from property groups, engagement scores built from event groups, and combined scores, for contacts, companies and deals. Optional decay cuts points by a set percentage after a set period, and threshold labels (A–C fit, 1–3 engagement) can show a record as C1: low fit, highly engaged. AI-built contact lead scores are listed for Marketing Hub Enterprise. Separately, HubSpot’s intent signals use research-intent and reverse-IP visit data for tracked companies, and tracking companies consumes HubSpot Credits. Check the current pricing page before you plan around any tier.
Zoho CRM uses Scoring Rules, found under Setup > Automation, for Leads, Contacts, Accounts and Deals. Field rules add or subtract points from record data; touchpoint rules score emails, calls and surveys when the related integration is enabled. Zoho’s Scoring Rules API documentation notes that signal rules work only on people-based modules such as Leads and Contacts. For predictive scoring, Zia Scores are available in Professional and above editions and need at least 200 records, with at least 75 records each for ideal and non-ideal conditions in the training data. As of November 2026, check Zoho’s current pricing and edition pages for limits. Our guides to HubSpot CRM and Zoho CRM cover the wider platforms.
In either tool the score is just a property on the record; routing, alerts or nurture based on it are configured separately in each platform’s automation features.
Which kind of lead scoring model should you start with?
Few closed-won deals or messy lifecycle stages? Fix CRM hygiene and use a fit-only grade.
Grade every lead A–C against the ICP. This alone stops sales chasing poor-fit leads.
Once intent actions are tracked reliably, add a points model with caps, negative scores and decay.
Back-test, then track conversion by band for a full sales cycle.
With consistent history, run a predictive score in parallel and compare bands.
How should marketing and sales govern the scoring model?
Treat the model as a shared agreement: document MQL and SQL definitions, the points table, thresholds, follow-up service level and who can change what. Give sales a one-click rejection reason, the fastest feedback loop you have, and review band conversion monthly and the model quarterly. Bargaon’s CRM service covers scoring fields and reporting. If you are setting up or rebuilding the CRM itself, plan scoring fields and lifecycle stages as part of your CRM implementation rather than bolting them on afterwards.
What are the most common lead scoring mistakes?
1. Rewarding email opens
What: Giving points for opens. Why: Apple introduced Mail Privacy Protection in 2021; Apple says it helps users prevent senders from knowing when they open an email and stops senders using invisible pixels to collect information, so opens recorded for Apple Mail users are unreliable. How: Score clicks, replies and on-site actions instead, and give opens zero or minimal points.
2. One blended number
What: Mixing fit and engagement into one total. Why: An engaged poor-fit lead looks like a quiet perfect-fit one. How: Keep separate fit and engagement scores or A1–C3 labels.
3. Scoring without outcomes
What: Points chosen in a workshop and never checked. Why: Untested assumptions persist for years. How: Back-test against closed-won and closed-lost data.
4. No decay or negative scores
What: Scores that only go up. Why: Old activity and disqualifying traits keep leads in the sales queue. How: Add decay to engagement and negative points for clear disqualifiers.
5. Thresholds that ignore capacity
What: An MQL threshold that ignores sales headcount. Why: Overloaded reps stop trusting MQLs. How: Size it to the volume the team can follow up on time.
6. Ignoring privacy
What: Scoring people on data they did not expect you to collect or use. Why: Data protection laws, including India’s framework under the Digital Personal Data Protection Act, which is being phased in, set rules on consent, purpose and notice. How: Explain profiling in your privacy notice, collect only what the model needs, respect opt-outs and get legal advice for your markets.
What does a 30/60/90-day lead scoring plan look like?
- Days 1–30: Agree MQL and SQL definitions, clean lifecycle stages, export 12 months of leads with outcomes and calculate lift.
- Days 31–60: Build the first points model with caps, negative scores and decay, back-test it, set thresholds against sales capacity, and add a rejection-reason field.
- Days 61–90: Go live, report conversion by band weekly and hold the first review to adjust weights.
Which related guides should you read next?
- Lead Management and Sales Hand-offs for routing, service levels and what happens after a lead becomes an MQL.
- Lead Generation and Qualification for the wider qualification process that scoring supports.
- HubSpot CRM for setting up properties, lifecycle stages and reporting in HubSpot.
- Zoho CRM for modules, Scoring Rules and Zia features in Zoho.
- Account-Based Marketing for applying account-level scores to target-account programmes.
Frequently asked questions
What is a good lead score?
There is no universal good score. A good score sits in a band your own data shows converts well enough to justify sales time.
How many criteria should a lead scoring model have?
Start small, perhaps eight to fifteen signals across fit, intent and engagement; fewer tested criteria beat dozens of minor ones.
What is the difference between an MQL and an SQL?
An MQL meets the agreed score threshold and is passed to sales. An SQL has been accepted by sales after a conversation confirms need, authority and timing.
Should existing customers be scored?
Usually not in the new-business model; route them to account management or a separate health model.
Is predictive lead scoring better than manual scoring?
Only with enough clean historical outcomes, and it still needs validating by score band.
How often should a lead scoring model be reviewed?
Check conversion by score band monthly and update the model quarterly, or sooner after a change to pricing, offer or target market.
Sources and further reading
- HubSpot Knowledge Base: Build lead scores to qualify contacts, companies, and deals (updated August 2026)
- HubSpot Knowledge Base: Build contact lead scores with AI (updated August 2026)
- HubSpot Knowledge Base: Use intent signals (updated September 2026)
- Zoho CRM: Scoring Rules how-to video
- Zoho CRM Developer: Create Scoring Rules API (v8)
- Zoho Help: Scoring Rules, Zia Scores
- Gartner: Sales survey finds 67% of B2B buyers prefer a rep-free experience (March 2026)
- Gartner: Survey finds 69% of B2B buyers turn to sales reps to validate AI-generated insights (May 2026)
- Apple Newsroom: Apple advances its privacy leadership with iOS 15 (June 2021)
- MeitY: Data protection framework
- Editorial note: the worked example, the score-band table, the conversion-by-band chart and the Bargaon Scoring Readiness Model are Bargaon editorial models with hypothetical values, not research data or client results.
Need a lead scoring model your sales team will trust? Bargaon’s CRM consulting service helps you define fit, intent and engagement criteria, set up scoring properties and reporting in your CRM, and validate the model against your own conversion data.