# Lead scoring: how to build a score your sales team will actually use

Migaru, Nov 19, 2026 · Guides

> What lead scoring is, fit versus engagement, rules-based versus predictive models, how to build one step by step, and why most scores end up ignored.

**Lead scoring** ranks leads, or whole accounts, by how likely they are to buy, so a sales team spends its time where it is most likely to pay off. A score usually combines **fit** (how closely they match the customers you win) with **engagement** (what they have done lately, such as meetings, replies and open deals), and subtracts for warning signs like silence.

Building a score is easy. Building one a team trusts is not: a number nobody can explain is a number people learn to ignore. The difference is almost always whether the score shows its reasons.

## Fit, engagement and intent

- **Fit:** who they are. Industry, company size, country, the technology they use, the role of the person. Fit says whether they could be a customer.
- **Engagement:** what they are doing with you. Meetings booked, replies, new people joining the conversation, an open deal. Engagement says whether they are becoming one.
- **Intent and outside signals:** what is happening at their company. A funding round, a hiring push, a new leader. These say whether now is a good time. [Buying signals](https://www.migaru.ai/blog/buying-signals) covers them in detail.

A strong score uses all three, and keeps them visible separately: a perfect-fit company that hasn’t replied in months needs a different next step from a poor-fit one that books a meeting every week.

## Rules-based or predictive?

- **Rules-based scoring** adds and subtracts points for signals you choose. It is transparent, easy to change and works from day one, but it is only as good as your guesses about what matters.
- **Predictive scoring** trains a model on past wins and losses to weigh the signals for you. It can find patterns people miss, but it needs a lot of clean history, and when it disagrees with a rep, it often can’t say why.

Most B2B teams under a few hundred closed deals a year are better served by rules they can read. Start there, and check the rules against outcomes every quarter.

## How to build a lead scoring model

- **Start from your won deals.** List what your best customers had in common before they bought: their size, industry, the role of the first person you spoke to, how the conversation started.
- **Choose a handful of fit criteria** from that list, the ones that genuinely separate winners from losers.
- **Choose engagement signals** that come before a purchase: a meeting booked, a reply from a senior person, a second person joining the thread, an open deal.
- **Add negative signals.** Silence past your normal follow-up time, a thread waiting on your reply, a deal stuck in one stage, a promise overdue, a champion who left. A score that only goes up is a score of how long you’ve known them.
- **Weight them,** with the strongest signals worth the most points, and keep the arithmetic simple enough to explain.
- **Let signals expire.** A meeting last week means more than one last year. Work out each score from what is true now.
- **Attach an action to each range.** A score that changes nothing about anyone’s day is decoration.
- **Review against outcomes.** Each quarter, check whether high scores won more often than low ones, and change the rules that didn’t predict anything.

## An example scoring model

An illustration to adapt, not a recommendation. Each account starts at 50 and stays between 1 and 99:

- Matches your ideal customer profile closely: **+10**
- Open deal: **+10**
- Meeting booked in the last two weeks: **+8**
- A new person from their side joined the conversation: **+5**
- Someone on your team knows them well: **+5**
- Raised funding or started hiring fast in the last month: **+5**
- No contact for longer than your follow-up time: **−10**
- They wrote last and are waiting on you: **−5**
- Deal stuck in one stage for three weeks: **−5**
- Your main contact has left the company: **−10**

Whatever the numbers, the record should list each signal behind a score, with its points and the fact it came from, so “why is this 78?” has an answer.

## Lead scoring or account scoring?

In B2B, several people at a company are usually involved in a purchase, and the one who filled in a form is rarely the one who signs. Scoring the account, with the activity of everyone there rolled up, gives a truer picture than scoring each person separately. Keep contact scores too, for deciding who at the account to talk to first.

## Why lead scores get ignored

- **Nobody can explain them.** Reps won’t follow a number they can’t argue with.
- **They only go up.** Without negative signals, every account looks warmer the longer you’ve known it.
- **They’re stale.** A score worked out monthly misses the reply that came in this morning.
- **They’re built on bad data.** If meetings and emails aren’t captured, engagement can’t be scored. [CRM hygiene](https://www.migaru.ai/blog/crm-hygiene) comes first.
- **Nothing happens at any score.** No queue, no alert, no change to who calls whom.

## How scoring works in Migaru

Every account and contact in Migaru starts at 50 and moves between 1 and 99 with each signal. An open deal, a booked meeting, new people joining the conversation or a strong connection with your team add points. Silence longer than your follow-up time (a month unless you set it anywhere from a week to three months), a thread waiting on your reply, a deal stuck in one stage for three weeks since it last moved, or an overdue commitment take them away.

No model is involved. The record page lists every signal with its points and the evidence behind it, and scores refresh about twice a day from your captured email, meetings and calls, so they move when the relationship does. Scores are on every plan. On Growth, companies with a website also score on fit with your ideal customer and outside signals such as a recent funding round, faster hiring or rising IT spend.

[AI pipeline generation](https://www.migaru.ai/product/ai-pipeline-generation) covers scores alongside warm paths and prospecting, and [the ideal customer profile](https://www.migaru.ai/blog/ideal-customer-profile) how to describe who you sell to.

## Frequently asked questions

### What is lead scoring?

A way of ranking leads or accounts by how likely they are to buy, usually by adding points for fit and engagement and subtracting them for warning signs, so a sales team knows where to spend its time.

### What is the difference between fit and engagement in lead scoring?

Fit is who they are: industry, size, location and role, compared with the customers you win. Engagement is what they are doing: meetings, replies, new people joining the conversation, an open deal. A useful score keeps both visible.

### Is predictive lead scoring better than rules-based scoring?

Not automatically. Predictive scoring needs a lot of clean history and often can’t explain itself. Rules-based scoring is transparent and works from day one, which is why many smaller B2B teams get more use from it.

### What are negative signals in lead scoring?

Signals that lower a score: no contact for longer than usual, a thread waiting on your reply, a deal stuck in one stage, an overdue promise, or a key contact leaving. Without them, scores only ever go up.

### How often should a lead score be updated?

At least daily, from activity captured automatically. A score worked out weekly or monthly misses the replies and meetings that should change who gets called today.

---

Canonical page: https://www.migaru.ai/blog/lead-scoring
Last updated: 2026-11-19
