B2B lead scoring: how to build a model your sales team will actually trust

Lead scoring B2B is really two scores

The most common mistake in lead scoring B2B teams make is adding everything into one total. A tiny company that opens every email and a perfect-fit enterprise that has done nothing yet can land on the same number, and those are completely different leads. To automate the pass or fail decision on top of a score, see automated lead qualification.

So split it:

• Fit score: company size, industry, geography, tech or tools they use, and the contact's role. This barely changes week to week.
• Intent score: page visits, replies, demo requests, documents read, meetings booked. This moves daily and should fade over time.

Then act on the combination. High fit and high intent goes to a rep today. High fit and low intent goes into nurture or a patient outbound sequence. Low fit and high intent is the interesting one: usually a student, a competitor or a curious freelancer, and worth a quick look before you bin it.

A simple grid of A, B and C is plenty to start. I'd rather have three buckets people use than a 0 to 100 scale they don't.

Start with ten rules, not forty

You don't need machine learning or a platform with predictive in the name. You need a short list of rules you can explain out loud in a meeting.

Pick five fit rules and five intent rules. For fit, ask your sales team what their best customers have in common, then check your closed-won list to see if they're right. For intent, list the actions that have historically come before a conversation.

Give each rule a small number of points and write down why it's there. If you can't say why a rule exists, delete it. A rule nobody can defend is a rule reps will distrust the first time it gives a strange result.

The numbers don't need to be clever. Mine have usually started as educated guesses and been corrected later. That's fine, as long as you plan to correct them.

Not all engagement is worth the same

This is where models fall over. Email opens are close to noise now, because privacy features and link scanners can trigger them with nobody there. I wouldn't give an open much weight at all.

A rough order of what I'd trust, from weakest to strongest:

1. Opens and generic clicks
2. Content downloads
3. Repeat visits to the site
4. Pricing, integration or comparison page visits
5. A reply, a booked call or a demo request

The pattern is that the closer an action is to evaluating a purchase, the more it should count. Reading a blog post is curiosity. Looking at pricing three times is shopping.

If you want to see how this plays out for a specific market, I wrote up the healthcare version of lead scoring, where buying groups are large and cycles are long. Everything in this post is the general model, and that one is what happens when you stress-test it against a difficult buyer.

Score the account when several people show up

B2B deals involve more than one person. If you only score individuals, you miss the clearest signal there is: three people from the same company all looking at your material in the same fortnight.

You don't need an enterprise setup for this. If your CRM can group contacts by company, add a rule that gives a bonus when two or more people from one account are active within thirty days. If it can't, a weekly manual look at which companies have multiple engaged contacts will still beat scoring each lead alone.

Document engagement is the signal most models skip

Think about what actually happens before a B2B deal moves. Someone sends a proposal, a deck or a one-pager, and then waits. Most scoring models only see the email open, which tells you almost nothing about whether the document was read.

When I was running growth at a data company, I sent a lot of decks, and the honest truth is that "opened" and "read" were very different outcomes that my tools treated as the same thing. A contact who spends time on the pricing page of your proposal and comes back to it the next day is telling you something no form fill ever will.

That's the part LiveDocument is built for. You send a PDF or an image file as a single link, record a short video walkthrough so the reader hears you explain it, and see page-level analytics on which pages they spent time on and whether they returned. It isn't a scoring tool and it doesn't replace your CRM. You'd feed what you learn into your own model, either by hand or through whatever you already use. But it gives you an intent signal that is stronger than an open, and most models don't have one at all.

Decay old activity and subtract for bad signals

A score that only goes up becomes useless. After a few months everyone looks warm.

Two fixes:

• Decay: reduce intent points as they age. A demo request from last week should count far more than one from last spring. Thirty to ninety days is a common window, but pick one that matches your sales cycle.
• Negative scoring: take points off for signals that mean "not now". Personal email addresses on a B2B product, competitors, students, roles that never buy, and leads who've gone silent after a long sequence.

I'd also add one hard rule: if a lead requests a demo or asks for pricing, it goes to a human regardless of score. Models are for prioritising the quiet middle. They shouldn't stand between you and someone who's raised their hand.

Check it against deals you've already closed

Before you trust the model, test it. Take your last dozen won deals and a dozen lost or stalled ones. Run them through your rules as they would have looked a month before the opportunity opened. If the wins don't score noticeably higher, the weights are wrong and you should change them until they are.

Then do it again every quarter. Markets shift, your product shifts, and a model built on last year's best customers will quietly start pointing at the wrong people.

The other check is the one that matters most: ask the reps. If they're ignoring a score band, find out why. They're usually right about something the data hasn't caught yet.

Where lead scoring fits with the rest of your process

Scoring is a prioritisation tool, not a strategy. It helps you decide who gets a call first. It doesn't help if the follow-up is slow, so it's worth pairing with a decent lead follow-up system, and it only works if the leads coming in are the right ones to begin with. If you're still working out your sourcing, how to prospect and qualify is the step before this one.

For small teams, you may not need a score at all yet. A plain A, B, C, as I covered in organising sales leads, does most of the job until you've got enough volume that people can't hold it all in their heads.

FAQ