Lead scoring

A 35-person cybersecurity company publishes a SOC 2 guide and gets 600 new leads. The marketing team wants to send them all to sales. That’s where lead scoring earns its keep: it ranks prospects by fit, behavior, and timing so reps know who deserves attention first.

The direct answer is simple. A useful score combines evidence that a company can buy with evidence that it might buy now. A content download alone usually isn’t enough.

The mistake most teams make with lead scoring

A common scorecard gives 10 points for an ebook download, 15 for a webinar, 5 for an email click, and 20 for a pricing page visit.

It looks tidy. It’s also easy to game by accident.

A marketing consultant can download three ebooks and click two emails while researching a client project. They might score higher than an IT director at a 700-person fintech who visits the pricing page once, reads the implementation page, and leaves.

The consultant has activity but no buying fit. The IT director may have both.

Teams get this wrong because they treat engagement as intent. It isn’t. Engagement tells you someone did something. It doesn’t tell you whether they own the problem, have budget, or work at a company you can serve.

My view: fit should usually beat activity. If a lead doesn’t match the market you sell to, no amount of email clicking should push them into an SDR’s queue.

What a score should measure

A score normally pulls from four kinds of evidence.

Firmographic fit covers company size, industry, geography, revenue, and sometimes the technology already in use. Role fit asks whether the contact has influence over the problem. A director of security is more relevant to a security platform than an intern, even if both download the same guide.

Behavior includes visits to product, pricing, and integration pages, along with demo requests and event attendance. Timing and intent come from events outside your website: a funding round, a new executive, an acquisition, a compliance project, or a technology change.

Some information is explicit. The prospect gives you their employee count on a form. Some is inferred. Several people from the same account visit an integration page during the same week.

That distinction matters when you define your ideal customer profile. A vague ICP produces a vague score. “Growing B2B companies” isn’t enough. “UK fintechs with 200 to 2,000 employees, a payments team, and a recent processor change” gives the model something useful to work with.

Build the model from a sales decision

Don’t start by choosing a scale from 1 to 100. Start with the decision the score is supposed to support.

Ask: what does an SDR need to see before spending ten minutes researching an account and contacting someone?

For one B2B software company, the answer might be a target industry, the right employee range, a contact who owns the relevant function, and a recent business event. The points come after that.

A rule-based model could give 30 points for the target industry, 20 for company size, 20 for a director or executive role, 25 for a demo request, and 15 for a pricing page visit in the last two weeks. It might subtract 30 for an unsupported market and 25 for a personal email address with no matching company.

The exact numbers don’t matter at first. The handoff does.

At one company, a score above 70 might create an SDR task to research the account. A score between 40 and 70 might keep the lead in nurture. Below 40, sales takes no action unless a rep finds a separate trigger.

A high score isn’t permission to send a generic sequence. It’s a reason to inspect the account, find the event behind the activity, and decide whether the message has a credible point. That’s where the SDR process either gets better or gets clogged.

A practical example

Consider a 60-person payments infrastructure company selling reconciliation software to fintechs and online marketplaces. It has eight SDRs and three account executives. The old model rewards nearly every interaction, so the CRM is full of “hot” leads that never become opportunities.

The team rebuilds the model around its actual customers. A fintech or marketplace with 200 to 2,000 employees gets 25 points. A finance, operations, or payments leader gets 20. A newly announced processor or bank-partner change adds 20. A public audit finding or compliance project adds 15. Two visits to the pricing or integration page within 14 days add 10. A demo or technical consultation request adds 25.

The team also subtracts 30 points for an unsupported market and 25 for a student, consultant, or personal email with no company match.

Lead A is a marketing manager at a 12-person agency. They downloaded the reconciliation guide, attended a webinar, and opened four emails. The activity looks good, but the company and role don’t fit. After the negative adjustments, the lead scores 5 and goes into an educational sequence.

Lead B is the VP of finance at a 480-person marketplace. The company announced a new European payment processor last month. The VP visited the integration page twice and requested a technical consultation. The score reaches 100.

The SDR task for Lead B includes a research note: confirm whether the processor change created a reconciliation gap, then reference the public announcement in the first message. The score tells the rep where to look. It doesn’t write the message for them.

Rule-based, predictive, or account scoring?

Most smaller B2B teams should start with rule-based scoring. It’s visible, easy to challenge, and simple to change when sales finds a bad assumption.

Predictive scoring can help companies with high lead volume and clean historical data. It looks for patterns in CRM records and estimates which leads are likely to convert. But if employee counts are missing, industries are inconsistent, and opportunity stages mean different things to different reps, the model will produce precise-looking nonsense.

Account scoring makes more sense when several people influence a deal. Instead of judging one contact, the team looks at activity across the account. A CFO, operations director, and technical lead all researching the product is stronger evidence than one highly active contact at the same company.

None of these approaches repairs a weak ICP or an unclear sales process.

How to tell whether the model works

Don’t measure success by the number of leads crossing the threshold. That only tells you the threshold is being used.

Compare score bands with actual outcomes. Do high-scoring leads accept meetings more often? Do those meetings become opportunities? Do the opportunities close? Also check how quickly an SDR acts and how often sales rejects a lead for being too small, irrelevant, or too early.

Sales rejection is often the best source of model improvements. If reps repeatedly reject companies below 100 employees, increase the weight of company size or add a hard exclusion. If high scores come mostly from email opens but those leads never book meetings, reduce the value of email opens. They’re cheap signals.

Review the model monthly for the first few months, then at least quarterly. Remove signals that don’t predict pipeline. Add triggers that show up repeatedly in won deals.

And keep early-stage leads out of the active queue without treating them as worthless. A good-fit company may simply lack a current project. Nurture is the right destination until something changes.

Lead scoring for outbound

Scoring works for outbound too. It just scores accounts and buying signals instead of form fills.

Imagine a 120-person HR software company targeting UK businesses with 500 to 3,000 employees. A new chief people officer, an acquisition, a major hiring push, or a payroll system change could move an account higher in the research queue.

A recent acquisition might justify a tailored message about consolidating employee data. An account with no relevant trigger can stay in a lighter cold outreach campaign.

That’s the useful boundary: lead scoring tells sales where to spend attention. It doesn’t create a reason for the buyer to reply.