Lead scoring: what it is, how to do it and how to calibrate lead points

Lead scoring is the practice of giving each lead a number that says, at a glance, how much of the sales team attention they deserve right now. Instead of working the queue in the order leads arrive, the company adds points as a lead gets closer to the ideal customer and as they show interest, and uses that total to decide who to contact first.

The idea solves a simple problem: not every lead is worth the same effort. A long contact list hides a few ready to buy and many that are not yet. Scoring makes that difference visible, splits hot leads from warm and cold ones, and keeps a rep from spending the morning on someone who only downloaded an ebook out of curiosity.

Every serious lead scoring model combines two independent dimensions. Fit answers "who is this lead" and measures how well they match the ICP: industry, company size, job title, region, budget. Engagement answers "what does this lead do" and measures their actions: opening emails, visiting the pricing page, requesting a demo, returning to the site several times in the same week.

• High fit and high engagement: hot lead, top sales priority.
• High fit and low engagement: matches the ICP but has cooled off, a nurturing target.
• Low fit and high engagement: enthusiastic but off-profile, be careful not to waste the team.
• Low fit and low engagement: cold lead, stays at the back of the queue.

Crossing the two dimensions, rather than adding everything into a single number, avoids the classic mistake of a click-happy but totally off-profile lead jumping the queue ahead of an account that fits the product perfectly.

Building a model starts by defining the destination. List the fit attributes that describe your best customers and the engagement actions that usually precede a purchase, then assign weights to each signal according to its strength. Requesting a demo is worth far more than opening an email, so it should score far more.

• Pick the fit signals: job title, size, industry and other ICP data.
• Pick the engagement signals: pricing page visits, downloads, email opens, product usage.
• Assign weights: high-intent signals score more; weak signals score little.
• Set the thresholds: how many points turn a lead into an MQL or SQL and hand it to sales.

The first model is never perfect. It is a hypothesis about what predicts a purchase, and the score only gains value when it is reviewed against what actually happens in the funnel.

It is worth distinguishing two types of scoring. Explicit scoring uses data declared by the lead or enriched from external sources: job title, industry, company size. It is stable and relates to fit. Behavioral scoring uses what the lead does over time: pages visited, emails opened, product events. It is dynamic and relates to engagement.

Because behavior ages, a good model applies decay: points earned from a visit three months ago are worth less today than yesterday. Without decay, old leads accumulate points by inertia and keep looking hot long after they have cooled. Decay keeps the score faithful to current interest, not to a frozen history.

In self-serve products and free trials, the strongest signal is not the job title or the industry, it is usage. When someone activates a core feature, invites colleagues or hits the free plan limit, they show real intent in a way no form captures. That pattern gives rise to the Product Qualified Lead (PQL).

That is why, in PLG, the weight of product engagement tends to outweigh demographic fit. A user who has already felt the product value, even at a company outside the classic ICP, often closes faster than a perfect-profile lead who never got past sign-up. Lead scoring in a PLG has to read usage events, not just marketing clicks.

A lead scoring model is only worth what the data confirms. Calibration means looking back at the leads that became customers and asking whether the high score really preceded the close. If score-90 leads close at the same rate as score-40 ones, the model is separating nothing and needs new weights.

The indicators that reveal this are MQL-to-SQL conversion and the win rate by score band: in a healthy model, the higher the score, the higher the advance and close rate. Funnel conversion benchmarks, such as those compiled by Benchmarkit, help compare your numbers with the market. The rule is to treat the model as alive: review the weights every quarter as the profile of who closes shifts.