B2B Lead Scoring That Sales Will Actually Trust

A B2B lead score should prioritize work, not pretend to predict a sale precisely. Separate company fit from evidence of buying interest, use a small number of observable inputs, and calibrate the thresholds against real sales outcomes. A score of 87 is meaningless if a salesperson cannot explain why it is higher than 62.

Many systems start by adding points for activity: opening emails, visiting a page, downloading a guide. The result can be a highly engaged student, competitor or job seeker above a quiet but ideal buyer. Fit and behavior must remain distinguishable.

Use two axes, not one mysterious number

Fit asks whether the company and role match the customers you can serve profitably. Factors might include industry, size, market, technical environment and role responsibility. Interest or timing asks whether something suggests an active evaluation: a relevant inbound request, a reply naming the problem, a procurement project, or a verified change at the account.

Keep negative rules outside the positive score. A current customer, competitor, prohibited market, student, personal email unrelated to the business, or company below your minimum viable size may be excluded regardless of page views. The ICP guide provides the fit logic; our buying-signals guide helps distinguish stronger timing evidence.

Salesforce's lead qualification training describes assigning point values to interactions after sales and marketing agree on qualification criteria. The useful lesson is the sequence: define a good prospect with sales first, then score behavior. The particular point values are a local business choice, not an industry standard.

A transparent example model

Consider an illustrative model. Fit grade A requires the right market, target industry, viable company size and relevant role. Grade B misses one preferred condition but is still serviceable; grade C falls outside the target. Interest level 1 is no evidence beyond a name in the database. Level 2 is a relevant content request or an observable trigger. Level 3 is a direct question about the problem, timeline or implementation.

The operating matrix is simple: A3 receives prompt human follow-up; A2 receives a tailored investigation; B3 receives a fit check before a meeting; C3 is reviewed for a possible new segment rather than automatically passed to sales; A1 remains in an appropriate nurture or research pool. The model tells the team what action to take, which is more valuable than a finely graded total.

Do not turn an open pixel into strong intent. Apple's Mail Privacy Protection prevents senders from observing some human opens reliably. A reply that states a problem carries different evidence from a tracking event.

Validate and recalibrate the score

Take the last several cohorts with enough time to progress. For each score band, compare sales acceptance, qualified meetings held, opportunities, wins and disqualification reasons. If A2 leads rarely become opportunities but B3 leads do, the weights or thresholds are wrong. If sales rejects “high scores” because the companies cannot buy, strengthen negative fit criteria.

Check selection bias. If sales only contacted high-scoring leads, the low-score group has fewer opportunities partly because nobody worked it. Run a small, controlled review of lower bands before declaring the model accurate. Record the version of the scoring rule, since changing weights mid-quarter makes historical comparisons hard.

The score also needs a decay rule. A three-month-old project signal is not equal to a reply from yesterday. Rather than an elaborate mathematical decay curve, require a recent date and revalidation for signals whose relevance expires quickly. If a person changes employers, the lead and account relationship should be updated.

Frequently asked questions

How many fields should the first model use?

Use only fields that can change a decision and that your team can verify. A simple matrix of fit and interest is enough to start; add complexity only when outcomes show it is needed.

Is a content download an MQL?

Not by itself. Define MQL using both fit and behavior, and agree with sales on the handoff. Our MQL-to-SQL guide gives the stage definitions.

Should a sales rep be allowed to override a score?

Yes, with a reason captured in the CRM. Overrides reveal missing signals or bad assumptions that can improve the model.

The next step

Ask sales to explain three high-scoring leads they accepted and three they rejected. If the score cannot explain that difference, simplify it and retrain on the actual decisions. Leadsify helps teams focus prospecting on accounts that can become qualified opportunities. Discuss your lead criteria.