A lead score decides which lead sales picks up first, and an SLA makes sure someone actually does it, on time. If you only build the first, you end up with a decorative number in the CRM. In this post I explain how I set up both together and how that translates into HubSpot and Salesforce.
What a useful lead score is (and isn’t)
A lead score is a number that answers two questions: does this person fit the customer we’re after, and are they showing interest right now? It’s useful when marketing and sales agree on which score turns a contact into an MQL and what happens next. It isn’t useful when marketing builds it alone, awards points for opening emails and never checks it against closed deals.
My starting rule: scoring doesn’t predict the future, it orders the work queue. If the score doesn’t change how sales prioritises, you don’t need it.
Separate fit from interest
The most common mistake is adding everything into one number. A student who downloads five ebooks can outscore an operations director who visits the pricing page once. That’s why I prefer two scores:
- Fit: job title, company size, industry, country. It changes little and comes from contact and company data.
- Engagement: pricing page visits, demo requests, email replies, event attendance. It changes fast, so it should decay over time.
- Negative points: free email domains, unsubscribes, profiles outside your market.
With two axes you can decide better: high fit and high interest goes to sales today; high fit and low interest gets nurtured; low fit and high interest is reviewed before spending sales time.
Write the SLA down
A marketing-sales SLA is a mutual commitment: marketing hands over leads that meet an agreed definition, and sales works them within a set time and records the outcome. Speed matters. A Harvard Business Review study of more than 2,000 companies, cited in this LeanData analysis, found that contacting a lead within the first hour made a meaningful conversation with a decision maker seven times more likely than waiting one more hour. And in the Blazeo benchmark covered in the same article, companies with a defined SLA responded within 15 minutes 54.9% of the time, versus 29.5% for those without one.
A reasonable starting point, which you then tune with your own data:
| Lead type | Criteria | Contact window | Owner |
|---|---|---|---|
| Demo or contact request | Direct action, any score | Same working day, ideally minutes | Sales |
| MQL with good fit | Fit and interest above threshold | 24 working hours | Sales (SDR) |
| High interest, doubtful fit | High interest, low fit | Review within 48 hours | Marketing and sales |
| Everyone else | Below threshold | No sales contact | Marketing (nurturing) |
The exact numbers matter less than three things: they’re written down, each row has an owner, and the CRM records what happened to every lead (accepted, rejected with a reason, converted).
How it works in HubSpot
In HubSpot, manual scoring lives under Marketing > Lead Scoring, where you can build a fit score from properties, an engagement score from events, subtract points with negative criteria and turn on score decay. Availability depends on your plan, so check it before you design anything. Predictive scoring, which learns from your won contacts, is limited to the top tiers and, according to this guide, works best with around 200 closed customers and clean data. If you have fewer, a well-thought-out manual model is usually more reliable.
For the SLA, a workflow that assigns the owner, creates a task with a due date and alerts someone if the lead is still untouched when the window expires is enough. Add a property for the rejection reason so you can measure it.
How it works in Salesforce
With Marketing Cloud Account Engagement (formerly Pardot) the separation is built in: score measures activity and grade measures fit. You can combine them in automation rules to pass a prospect to sales when both clear the threshold. Then, in the CRM, use lead queues with assignment rules, tasks with due dates and a report of leads with no activity past the window. The principle is the same as in HubSpot: assign, set a deadline, alert and measure.
Common mistakes
- Giving lots of points to low-value actions such as opening an email.
- Letting interest never decay, so a year-old lead still looks hot.
- Building the model without sales, so nobody trusts it.
- An SLA with no follow-up: if nobody looks at the report, it isn’t an SLA.
- Not recording rejections with a reason, and losing the information that improves the model.
A quarterly review in four steps
- Pull last quarter’s won customers and check what score they had when they became MQLs.
- Compare conversion rate by score band: if the top band doesn’t convert better than the middle one, the model isn’t ranking anything.
- Check the share of leads contacted within the window and the most frequent rejection reasons.
- Adjust weights and thresholds with sales, and note what you changed and why.
Frequently asked questions
How many points should an MQL have?
There’s no universal number. Set it by looking at the score of the leads that ended up buying, and tune it with sales until the MQL queue is workable.
Manual or predictive scoring?
Start with manual: it forces you to define your ideal customer and it’s easy to explain to sales. Predictive makes sense once you have enough customers and reliable data.
What response SLA is reasonable?
For direct actions like a demo request, as fast as possible, in minutes if you can. For other MQLs, one working day is usually realistic; what matters is meeting it and measuring it.
If you already run something similar in your team, I’d love to hear how you do it in the comments or through the contact page.
