Lead scoring for B2B outbound: a simple model you can use today

What lead scoring is, and why outbound needs a different version

Lead scoring gives each lead a number so you can decide who to contact first. Most guides describe scoring inbound leads: someone visits your pricing page, downloads a guide, attends a webinar. That behaviour is the engagement score.

An outbound list has no behaviour yet. You chose these people, they did not choose you. So the score has to lean on two things you do have: fit, which is how closely the person and company match your ideal customer profile, and signals, which are facts that suggest timing or interest.

Deep dive: How to build an ideal customer profile

The rule that keeps scoring honest: fit is the gate

Score fit and signals separately. A lead that does not fit your ICP should never be promoted by a signal. A competitor's follower who is a student, or a commenter at a company ten times too small, is not a good lead however warm they look.

Use fit as a minimum threshold. Inside the leads that pass it, use signals to decide the order.

Build the fit score in five steps

1. 1. Start from your ICP

   Turn each ICP criterion into a scoring attribute: role and seniority, company size, industry, location. If a trait is not in your ICP, do not score it.

2. 2. Give each attribute points

   Weight what predicts a win more heavily. Role and company size usually matter more than location. A total of 100 points keeps it easy to read.

3. 3. Add negative points

   Subtract for traits that predict a waste of time: a personal email address, a very junior title, a company far outside your size range.

4. 4. Set a minimum fit

   Choose a threshold below which a lead is not contacted at all. Start generous and tighten as you learn.

5. 5. Calibrate on real outcomes

   Run your 20 best customers and 20 lost or poor-fit deals through the score. If it cannot tell the two groups apart, the weights are wrong. Change them and test again.

A worked 100-point model

This is an example for a company selling to sales leaders at software firms. The weights are made up to show the mechanics.

• Role: 30 points for the buyer title (VP of Sales), 15 for an adjacent title (sales operations manager), 0 otherwise.
• Company size: 20 points for 50 to 500 employees, 10 for 11 to 49, 0 outside that range.
• Industry: 20 points for software, 10 for adjacent industries, 0 otherwise.
• Location: 10 points for your target country.
• Verified work email: 10 points. Without one you cannot reach them by email.
• Signals, up to 10 points: for example 5 for following a competitor, 5 more for engaging with a competitor's post.
• Negative: minus 20 for a personal email domain, minus 30 for an intern or student title.

Scoring three example leads

Jordan Reyes, VP of Sales at a 120-person software company in the target country, with a verified email who follows a competitor and commented on its launch post: 30 + 20 + 20 + 10 + 10 + 10 = 100. Contact first, with a message that references the problem, not the competitor.

Elena Rossi, COO at a 40-person software company in the target country with no email found: the role is not the buyer (0), size 10, industry 20, location 10, no email 0, no signals 0, so 40. Hold until you find an email or a better reason to contact her.

Sam Okafor, founder of a 600-person logistics company, personal email: role 0, size 0, industry 0, location 10, minus 20 for the personal address: negative. Do not contact.

All three are fictional.

Turn scores into action

Put the score and the tier in a column next to each lead so the list can be sorted. Reps trust a score they can see being built from visible columns.

• Tier A (for example 80 and above): contact first, with a personalised first line and the full sequence.
• Tier B (50 to 79): standard sequence.
• Tier C (below 50): hold, or enrich further before deciding.
• Below your minimum fit: do not contact.

Keep the model from rotting

Once a simple rules-based model is stable and you have enough outcomes, you can try a model trained on your own history. Start simple.

• Review the weights every quarter against who actually replied and who bought.
• Refresh the data you score on. A stale job title makes a wrong score.
• Build it with the people who use it. If reps do not trust the scores they will ignore them.
• Do not score what you cannot act on. Every attribute should change what you do.
• Do not add signals just because they exist. Check that they predict replies.

Frequently asked questions

What is lead scoring?

A way of giving each lead a number so you can decide who to contact first, based on how well they fit your ideal customer and any signals of interest.

What is the difference between fit and engagement scoring?

Fit measures how well a person and company match your ICP. Engagement measures what they have done, such as visiting your site. Outbound lists usually have fit and a few signals, not engagement.

What score should count as a good lead?

There is no universal number. Choose a threshold that sends you a number of leads you can contact well, then adjust as you compare scores with real replies and wins.

How many criteria should a score have?

Fewer than you think. Five to eight that predict outcomes beat twenty that nobody understands.

Do I need predictive scoring?

Not at first. A simple rules-based model you understand is enough until you have a large history of outcomes to learn from.