Lead scoring is a method of ranking leads by assigning points for who they are and what they do, so sales teams spend time on the prospects most likely to buy and marketing knows when to hand a lead over. Points come from fit, such as role, company size and industry, and from engagement, such as visiting the pricing page, attending a demo or opening emails. This guide explains what lead scoring is, how to build a scoring model step by step, examples, predictive scoring, how to set thresholds and keep the model honest, and the mistakes to avoid.
Short answer: A lead score is a number that estimates how ready and how suitable a lead is. Combine a fit score and an engagement score, set a threshold for sales follow-up and review the model against real closed deals.
What is lead scoring?
Lead scoring turns a pile of leads into a ranked list. Without it, sales representatives either chase every lead, wasting time on poor fits, or pick leads by instinct, missing good ones. With it, a lead that matches your ideal customer and has recently shown strong buying signals rises to the top and is contacted quickly.
The score lives on the lead or contact record in your CRM and updates automatically as data and behaviour change. It is also a shared language between marketing and sales, which makes arguments about "lead quality" much more concrete.
How lead scoring works

- Capture data. Collect firmographic, demographic and behavioural data from forms, the website, email and the CRM.
- Assign points. Add points for positive signals and subtract them for negative ones.
- Combine into a score. Fit and engagement scores are summed or kept side by side.
- Apply thresholds. Scores above a set level mark a lead as marketing qualified or ready for sales.
- Route and act. Assign the lead, create a task and notify the owner.
- Review and tune. Compare scores with outcomes and adjust the weights.
Types of lead score
| Score | Based on | Example signals |
|---|---|---|
| Fit (explicit) | Who the lead is | Job title, company size, industry, country, technology used |
| Engagement (implicit) | What the lead does | Pricing page visits, demo requests, email clicks, webinar attendance, content downloads |
| Negative | Signals of poor fit or lost interest | Personal email address, student, competitor, unsubscribed, no activity for months |
| Predictive | Patterns learned by a model from past wins | Combinations of attributes that resemble customers who bought |
A lead scoring model example
| Attribute or action | Points |
|---|---|
| Job title is a decision maker | +15 |
| Company size within target range | +10 |
| Target industry | +10 |
| Visited pricing page | +15 |
| Requested a demo | +30 |
| Opened three or more emails in a month | +5 |
| Attended a webinar | +10 |
| Free email domain | -10 |
| No activity for 60 days | -15 |
| Competitor or student | -30 |
The numbers matter less than the ratios. A demo request should clearly outweigh an email open, and a bad-fit signal should be able to cancel interest from a lead you could never serve. Treat these points as a starting hypothesis to test, not as truth.
How to build a lead scoring model
- Agree the definition of a good lead. Marketing and sales must agree on the ideal customer profile and what "sales ready" means.
- Analyse closed-won deals. Look at the attributes and actions common to customers who bought, and to leads who did not.
- List scoring criteria. Choose a short set of fit and behaviour signals you can actually capture.
- Assign weights. Give more points to signals closer to buying and negative points to poor fits.
- Set thresholds. Choose a score for marketing qualified and one for sales ready, and test them against history.
- Build it in your CRM. Configure the scoring property and routing. See below for HubSpot and Salesforce.
- Pilot, then refine. Run for a quarter, compare to outcomes and adjust.
The best lead score is the one sales trusts. If representatives ignore it, it is not a model, it is a number.
Lead scoring in HubSpot and Salesforce
HubSpot offers lead scoring properties with positive and negative criteria in its paid tiers. Combine them with lifecycle stages and workflows to notify and assign when a threshold is crossed. In Salesforce, scoring can be built with formula fields, Flows or Einstein tools, and leads can arrive through web-to-lead. See our guides to Salesforce Flow Builder and Einstein AI. Marketing platforms such as Marketo and Account Engagement also include scoring.
Predictive lead scoring
Predictive scoring uses machine learning to find patterns in past conversions and score new leads by how closely they resemble customers. It can discover combinations that humans would not guess and removes some manual tuning. It needs a good volume of history, usually hundreds of won and lost outcomes, and clean data. It also needs explanation: sales will trust a score more when it shows the top reasons behind it. Many teams start with a rules-based model, then add predictive scoring once the data and process are stable. Our AI and machine learning services help with this step.
Lead scoring thresholds and handoff
| Stage | Typical meaning | Action |
|---|---|---|
| Subscriber / lead | Known contact, early interest | Nurture with content |
| Marketing qualified lead (MQL) | Fits the profile and shows meaningful interest | Accelerate nurture and review |
| Sales qualified lead (SQL) | Sales has confirmed need, authority and timing | Assign an opportunity |
Many teams score for two decisions: when a lead is ready for sales, and when a lead should stay in nurture. Leads that are a great fit but not yet engaged should flow into a lead nurturing programme rather than being discarded.
Lead scoring for B2B and B2C
B2B scoring leans on firmographics and the buying group. Company size, industry, role and the number of people from one account engaging all matter, so many teams score at the account level as well as the lead level. B2C scoring leans on behaviour and recency: browsing, basket activity, purchases and response to offers. The method is the same in both, but the signals and the speed of the cycle differ, so build separate models rather than forcing one to fit both.
Common lead scoring mistakes
- Building the model in marketing alone, without sales agreement.
- Counting every email open equally, although some are machine-generated.
- Scoring only activity, so a poor-fit visitor with heavy engagement looks like a hot lead.
- Never decaying old activity, so a lead looks warm because of something they did a year ago.
- Too many criteria that no one can explain.
- Setting the model once and never comparing it to results.
Practical tip: Add score decay. Subtract points for inactivity so the list reflects current interest, and review the closed-won leads each quarter to check that high scores really predicted wins.
How to measure lead scoring performance
- Conversion rate by score band: higher bands should convert much better.
- Sales acceptance rate of leads passed over.
- Time from threshold to first contact.
- Win rate and average deal size by score.
- Share of closed-won deals that scored below the threshold, which shows what the model misses.
Our CRM and lifecycle services include scoring design, CRM configuration and the lifecycle processes around it. Pair scoring with customer segmentation to tailor follow-up for each type of lead.
Key takeaways
- Lead scoring ranks leads by fit and engagement so sales focuses on the best opportunities.
- Use positive and negative criteria, and decay old activity.
- Agree the definition of a good lead with sales before building the model.
- Predictive scoring needs clean history and explainable outputs.
- Compare scores with real outcomes every quarter and adjust.
Quick glossary
| Term | Meaning |
|---|---|
| MQL | Marketing qualified lead |
| SQL | Sales qualified lead |
| ICP | Ideal customer profile |
| Firmographics | Company traits such as size and industry |
| Score decay | Reducing points over time for inactivity |
Read next: HubSpot CRM: Features, Hubs, Pricing and Setup Guide · HubSpot Operations Hub · HubSpot Campaigns
Frequently asked questions
What is lead scoring?
Lead scoring ranks leads by assigning points for fit and behaviour, so sales teams prioritise the prospects most likely to buy.
How do you create a lead scoring model?
Agree a definition of a good lead, analyse closed-won deals, choose fit and behaviour criteria, assign weights, set thresholds, build it in your CRM and refine it against outcomes.
What is a good lead score?
There is no universal number. A good score threshold is the level at which leads convert at a meaningfully higher rate, which you find by comparing scores with historical outcomes.
What is the difference between fit and engagement scoring?
Fit scoring rates how well a lead matches your ideal customer by attributes such as role and company size. Engagement scoring rates interest from behaviour such as page visits and demo requests.
What is predictive lead scoring?
Predictive scoring uses machine learning on past conversions to score new leads by similarity to customers who bought. It needs enough clean historical data.
What is the difference between an MQL and an SQL?
An MQL is a lead that marketing considers qualified by fit and engagement. An SQL is a lead that sales has accepted after confirming need, authority and timing.
How often should a lead scoring model be reviewed?
Review at least quarterly. Compare score bands with conversion and win rates, and adjust weights when your product, market or marketing channels change.
Can small businesses use lead scoring?
Yes. A simple model of five to ten criteria in a CRM such as HubSpot can help a small team prioritise follow-up without complex tools.
