How scoring works
Every finished lead gets a score from 0 to 100 — from your product's scored questions, or from Claude when there are none — plus the score your result page shows.
Productised scores people in two places, and it helps to keep them apart:
- The lead score — one number from 0 to 100 for each person who finishes your product. It's what you see on the Leads page, the dashboard's Latest leads, your product's Analytics and in exports, and it's what decides who counts as a hot lead.
- The result score — the score, tier and analysis your AI Agent writes for the person's own result page.
The lead score
Every lead that reaches the end of your product gets a lead score. It comes from one of two places.
1. Scored questions
When a product has scored questions, the lead score is calculated from the person's answers. Only scale, yes/no and multiple-choice questions can be scored. Each one has:
- Points per answer — for a choice or yes/no question, how many points each answer is worth (the strongest answer is worth 10, the weakest 0). For a scale, where the answer sits on the scale, with higher or lower set as the better end.
- A weight from 1 to 3 — how much the question matters. Each question is worth 10 points for every unit of weight.
The lead score is the points earned out of the points possible, as a percentage. A question the person didn't answer is left out, rather than counted as zero. The same answers always give the same score, so it's consistent and easy to explain.
2. The AI fit score
When a product has no scored questions, Claude scores the lead instead. When the conversation finishes, Claude reads the person's answers against what your product is trying to learn about them and gives:
- a score from 0 to 100 — how strong a lead they are, on fit, need and readiness to act, and
- a one-line reason for it.
These leads show a small AI marker next to their score on the Leads page. Hover over the score to read the reason. The reason is also included in your CSV export.
The AI fit score runs in the background just after the lead is saved, so it never slows down the person finishing your product. It uses your product's Claude key — your own key if you've added one — and isn't charged in credits.
Which leads get a score
- Only finished leads are scored. Someone who leaves part-way through has no score.
- A product with scored questions always uses them, even if the person's answers only cover some of them. If they answered none of the scored questions, Claude scores them instead.
- Leads that finished before the AI fit score was introduced keep the score they had, or stay unscored.
Low, medium, high and hot leads
Every scored lead also gets a rating:
| Rating | Lead score |
|---|---|
| Low | 0–39% |
| Medium | 40–69% |
| High | 70% and above |
These are the default cut-offs. A product's decision model can set its own.
A hot lead is one its product rates high. The Hot leads figure on the Leads page counts them, and the Hot only filter lists just those people.
A rating and its label (for example "High fit") are only given when there's a real score behind them. A lead without a score has no rating.
Turning scoring on
Scoring is set up when the product is built:
- Build with AI and building with Claude switch scoring on for every scorecard, assessment, diagnostic, audit, readiness check and quiz. They choose which questions to score, the points for each answer, and each question's weight.
- Advice, intake and booking products are built without scored questions. Their leads get an AI fit score instead.
On the canvas, open the AI Product node to see which questions are scored. Each one is marked Structured scoring enabled, with its weight.
The result score
Your result page can show a score of its own, written by the AI Agent. The agent reads the whole conversation and writes output fields such as overall_score, tier_label, dim1_score, dim1_insight and action_1. You set these up by describing what you want in plain language, and Claude wires them up (set_agent_fields).
Those fields fill your result page through {{ object:* }} tokens, for example {{ object:overall_score }} and {{ object:tier_label }}. Because the score is reasoned, two nearly identical people can get meaningfully different results when one detail changes the picture.
The result score is part of what the person sees. It doesn't change their lead score, which is what the Leads page, the dashboard and Analytics use.
Tiers on the result page
The agent's tier_label (for example "Ready", "Stabilising" or "Not yet") is the person's segment on their result page. To tailor the result by tier, tell the agent how each tier's message and call-to-action should differ. The single result page then adapts to the tier each person lands in.
What this replaces
Most quiz tools use conditional logic: hand-written if/then rules that route people through a fixed tree to one of a few pre-written outcomes. Productised doesn't. The AI Product node adapts its questions as the conversation goes, the lead score ranks your leads for you, and the AI Agent writes each person's result — so there's no branching tree to build or maintain.