# Lead scoring in Linkubit: signals, rules in words, and what a score sets off

> How a contact's score is built from rules over what they did and what they told you, when it is recomputed, and how a score qualifies a lead.

Source: https://www.linkubit.com/docs/lead-scoring-explained  
Section: Linkubit help centre › How the AI agent works  
Updated: 2026-10-09

## In short

A score is the sum of every active scoring rule whose condition holds, and each contribution is listed on the contact. Rules test signals (messages, goals, products sent, payments), your fields, or a sentence the agent judges as it reads (Pro and up). A score change fires contact.scored — and on the starter workflows, a score of 40 or more qualifies the lead.

## What a rule can test

| Group | Examples |
| --- | --- |
| **Signals** — what the contact did, computed when the score is | Messages from them, Messages to them, Conversations, Channel used, Days since they last wrote, Goal completed, Goals completed, A goal is running, Product sent to them, Qualification class, Lead source, Replied to a campaign, Payment requested, Paid, Products they're interested in, Products they want |
| **Contact** | Lifecycle stage, Score, Country |
| **Your fields** | Everything in Fields you collect › About each customer, and company fields when the contact has no value of its own |
| **In words** (Pro and up) | "They are planning an event for more than 50 people." The agent answers it as it reads the conversation. |

Several conditions can be combined with AND and OR. **Points** run from −100 to 100; a negative rule pulls a lead down (the starter **Disqualified** rule is −100).

## Rules written in words

- Answered on the agent's own turn, in the same call that reads your fields — 0.1 credit per turn however many such rules you have, at most 10 active.
- Once true for a contact, it stays true, with the agent's reason shown. Editing the sentence clears every answer, because a changed question has no answer yet.
- Only an agent turn answers them: a conversation a person took over, or a fact typed on the contact, produces no answer. Use a structured rule for those.

## When the score is recomputed

After every agent turn and every field the agent writes, after a product or payment request is sent, when a payment settles, when someone edits the contact, and when a website enquiry arrives. Signals are computed at that moment, never stored, so a rule on "Messages from them ≥ 5" holds from the fifth message on. When the total changes, `contact.scored` fires — once per change, not once per recompute.

## The starter rules

| Rule | Points |
| --- | --- |
| Replied to us (1+ message) | +10 |
| Engaged — five or more messages | +15 |
| Named a product | +20 |
| Asked us to follow up (wants a product) | +25 |
| Was sent a product | +10 |
| Completed a goal | +20 |
| Qualified — high | +40 |
| Qualified — medium | +20 |
| Disqualified | −100 |

> **A score band is a qualification.** The starter workflow **Score qualifies the lead** qualifies at 40 (low), 60 (medium) and 80 (high). On the starter rules, a customer who replied (+10), sent five messages (+15) and gave one detail your own rules reward (+20) is at 45 — qualified as low, which fires `lead.qualified`. Raise the band, or make your own rules reward what actually qualifies, before hanging anything consequential on `lead.qualified`.

## Questions

**Why did the score jump?**

Open the contact: the breakdown lists every rule that contributed. The thread's automatic events show each change with the rule that caused it.

**Does the AI decide the score?**

No. The score is a sum of your rules. The AI only answers rules you wrote in words, and its reason is shown.

## Related
- [From a conversation to a lead to a deal](https://www.linkubit.com/docs/conversation-to-lead-to-deal)
- [Workflows, explained](https://www.linkubit.com/docs/workflows-explained)
- [Pipeline stages, scoring and workflows](https://www.linkubit.com/docs/pipeline-scoring-and-workflows)