Guides

How leads are scored, and what moves a score

Last updated 2026-09-14

Every lead carries a score out of 100 and a band — hot, warm or cold. It is worth understanding how it is built, because it decides which leads you are shown first and, if you have the outcome loop switched on, which ones Meta and Google are told to go and find more of.

It is mostly arithmetic on what the person actually told you, not a language model's impression of how keen they sounded.

Why it is not just the AI's opinion

Scoring used to be left entirely to the model, and the trouble with that is that an enthusiastic message scores well while a terse one that contains a budget, a location and a timeline scores badly. Warmth is not intent.

So the score is now computed from signals you can point at, with the model's read used only to raise the floor — it can rescue an obviously hot lead from being under-scored, but it cannot talk a lead up on charm alone.

The six fields that carry most of the weight

These are the qualification fields. The more of them the conversation has actually captured, the higher the lead scores:

Budget what they can spend.
Location where they want it.
Unit type the configuration — 3BHK, and so on.
Property type what kind of property.
Timeline when they intend to move or buy.
Purpose to live in, or to invest.

Together these are worth up to 55 points — the largest single component, and the reason the assistant asks qualifying questions rather than just answering.

How the score is actually built

Qualification completeness — up to 55 points. The share of those six fields that are filled. All six is the full 55; three of six is about 28.
Engagement — up to 30 points. How many messages the person actually exchanged, saturating at around eight. Beyond that, more messages do not raise the score — a long conversation is not the same as a keen buyer.
Appointment interest — 15 points. Wanting a visit is the strongest single signal a buyer gives, and it is worth a flat fifteen.
The model's floor. If the assistant reads the lead as hot, the score is lifted to at least 80; warm lifts it to at least 55. It can only raise, never lower.

Where hot, warm and cold begin

Hot 75 and above.
Warm 45 to 74.
Cold below 45.

A useful sanity check: someone who gives a budget, a location and a configuration, exchanges a handful of messages and asks about a visit lands comfortably in hot. Someone who sends four messages and reveals nothing stays cold, however friendly the exchange was.

What moves a score, and what does not

Moves it: the person answering a qualification question, or asking for a visit. That is essentially the whole list.

Does not move it: more messages past roughly eight, politeness, how fast they reply, or how long the conversation has been open. A lead does not decay to cold by sitting there, and it does not warm up by chatting.

The one to watch

A lead can be genuinely serious and score cold, because they asked two sharp questions, got their answers, and left to think. The score measures what was captured, not what the person intends. Treat cold as “we do not know enough yet”, not as “not a buyer” — that is what the follow-up sequence is for.

What to do with it

Work hot leads personally, let the assistant keep working warm ones, and let the follow-up sequence handle cold. The score is a queue, not a verdict.

It matters twice over if you advertise: with the outcome loop switched on, a qualified lead is reported back to Meta or Google with a value attached, which is what teaches the ad platform to go looking for buyers rather than for repliers. That is covered in Telling Meta and Google which leads were real.

Know which five leads to call today

Every enquiry qualified and scored automatically, so the list you work in the morning is ordered by who is actually ready.