Lead scoring gives each lead a number that estimates how likely they are to become a customer, so sales can call the best ones first. The number comes from two things: how well the lead fits your ideal customer, and how much they have engaged with you. Both HubSpot and Salesforce can calculate it, using rules you write or a model that learns from your past deals.
This guide explains the models, how to set it up in each CRM, and what usually goes wrong. For applying a model in your own CRM, see our lead scoring model application automation.
What is lead scoring?
It is a way of ranking leads so limited sales time goes where it is most likely to pay off. A lead who runs a company of the right size in your target sector and has visited your pricing page three times should rank above a student who downloaded one PDF.
A score is a prompt, not a verdict. It helps sales decide who to contact first and helps marketing decide when a lead is ready to hand over. See marketing qualified lead handoff to sales for how a threshold becomes a workflow.
Lead scoring models: fit vs engagement
Most scoring models combine two separate scores.
Fit describes who the lead is: job title, company size, industry, country, revenue. It changes rarely. A great fit who has not engaged is worth nurturing.
Engagement describes what the lead has done: visited pages, opened or clicked emails, filled in forms, attended an event, replied to sales. It changes constantly and should usually decay over time, since a click six months ago says little about today.
Keeping them separate is useful. High fit and high engagement means call now. High engagement with poor fit usually means a researcher, a competitor or a customer’s colleague.
Rule-based vs AI-learned scoring
Rule-based scoring means you assign points: +10 for a director title, +5 for a pricing page visit, -10 for a personal email address. It is transparent and easy to explain, and you can start with no historical data. The weakness is that the weights are guesses, and they go stale.
AI-learned (predictive) scoring trains a model on your past leads and which became customers, then finds the patterns itself. It can spot combinations you would not think of. It needs enough closed outcomes to learn from, and it is harder to explain to a sales rep who asks why a lead scored 82.
A common approach is to start with simple rules, record outcomes properly, and move to predictive scoring once you have the volume. Then tune it. See lead scoring model tuning.
Lead scoring in HubSpot
HubSpot’s lead scoring tool lets you build score properties. According to HubSpot’s documentation, there are fit scores (based on property values such as job title or company size), engagement scores (based on actions such as site visits or email clicks) and combined scores. You build them from Marketing, then Lead Scoring, and criteria are grouped into rules. HubSpot also describes an option to create scores with AI. Which score types and features you can use depends on your subscription, so check HubSpot’s current documentation for your setup. See our HubSpot tool page.
A sensible walkthrough:
- Decide what a good lead is, using your last year of closed-won deals.
- Create a fit score from the properties that distinguish them.
- Create an engagement score from the actions that preceded a sale.
- Set a threshold and use it in a workflow to assign an owner or create a task.
- Review monthly against what actually closed.
Lead scoring in Salesforce
Salesforce supports scoring in a few ways, and teams often combine them. You can write your own score with formula fields or automation, so you control every point. Salesforce also offers Einstein Lead Scoring, a predictive score that learns from your lead conversion history. Its availability depends on your Salesforce edition and licences, so confirm this with Salesforce. For marketing behaviour, Account Engagement (formerly Pardot) provides behavioural scoring and grading. See our Salesforce tool page.
In general terms: set up the fields you need, turn on or build the score, make sure it is visible on the lead record and in list views, and trigger assignment or follow-up from it. If you use Zoho instead, see Zoho CRM AI lead scoring.
What data do you need?
- Consistent lead source and status values.
- Contact and company fields filled in, especially those used for fit.
- Tracking of engagement: email, website, forms.
- A clear conversion definition, such as “opportunity created” or “deal won”.
- Closed-lost reasons, not only wins.
Our duplicate contact and lead merging automation helps with the usual clean-up before scoring.
Lead scoring pitfalls
Too few closed deals. A model trained on twenty wins finds noise. Use rules until you have more history.
Bad outcome data. If reps do not mark deals won or lost properly, predictive scoring learns the wrong lesson.
Bias. A model learns from past decisions. If you historically ignored a segment, it will score that segment low and confirm your own habit. Review scores for unintended patterns, especially if you use personal data. Avoid sensitive attributes.
No decay. Old activity inflates scores.
No feedback loop. If reps do not trust the score, they ignore it. Show the reasons behind a score and ask for their feedback.
For GDPR, scoring is profiling of individuals, so be transparent in your privacy notice and see GDPR and AI automation in Malta. For the wider CRM context, read CRM automation with HubSpot and Salesforce.
Frequently asked questions
What is lead scoring in simple terms?
It is a ranking system that gives each lead points based on who they are and what they have done, so sales know who to contact first.
How many closed deals do I need for AI lead scoring?
There is no single number, and it depends on the tool. As a rule, hundreds of examples with clear outcomes give a model something to learn from, while a few dozen rarely do. Check your CRM’s documentation for its minimum requirements.
Should I use HubSpot or Salesforce for lead scoring?
Use whichever CRM you already run. Both can do it. The quality of your data matters more than the platform.
How often should I update the scoring model?
Review it at least quarterly against closed deals, and sooner if your offer or target market changes.
Where Automata fits
We apply and tune scoring models inside your existing CRM, whether that is HubSpot, Salesforce or Zoho, and wire the results into assignment and follow-up. Start with lead scoring model application, and see lead scoring model tuning if you already have a model that has drifted.
If you want a second opinion on your current setup, contact us.