How AI Twin lead scoring works

Every AI Twin conversation is scored 0–100 on two axes: intent (does this visitor want to buy?) and fit (are they the right customer for you?). The combined score determines whether the conversation reaches your inbox in real time, gets pushed to your CRM, or stays quietly logged for later review.

Intent signals

• Explicit buying language — "how much", "can we start", "send a quote", "trial".
• Question specificity — asking about your integrations vs. asking what you do.
• Depth of the conversation — more turns typically means higher intent.
• Time on page before starting the conversation.
• Repeat visits within the same session.
• Filling out contact fields voluntarily.

Fit signals

• Work email domain — matches your ideal customer profile (ICP) industries and headcount bands.
• Firmographic enrichment — industry, headcount, region, tech stack (best-effort from public data).
• Referrer — visitors from a target account list rank higher.
• Language and geography — match against your served regions.

Define your ideal customer profile (ICP)

Settings → Leads → ICP. Describe the customer you want to prioritise:

• Industries (SaaS, e-commerce, agencies, etc.).
• Headcount bands (1–10, 11–50, 51–200, 200+).
• Regions.
• Optional target account list (upload CSV of company domains).

Conversations matching your ICP get a fit boost; those clearly outside get down-weighted.

Notification threshold

By default, conversations scoring above 70 land in your inbox in real time. Adjust in Settings → Leads → Notifications. Common patterns:

• Threshold 80 — noisy sites where only very-high intent should page a human.
• Threshold 50 — early-stage products where any warm lead is precious.
• Threshold 90 with CRM sync at 70 — the CRM captures more, sales gets fewer pings.

How the score is presented

Every conversation shows the score with a short explanation — which signals fired, and by how much. This is the same explanation that goes into the CRM sync so your salespeople know why a lead landed on their desk.

Best practices

• Tune the threshold weekly for the first month — you'll converge fast.
• Feed closed-won and closed-lost outcomes back into Reality Twin (via CRM sync) to improve scoring over time.
• Use a lower threshold for CRM sync than for real-time notifications.

Common mistakes

• Setting the threshold too low and ignoring notifications — nothing gets acted on.
• Not defining an ICP — every conversation ends up mid-scored.
• Trusting fit signals blindly — always read the transcript before calling.

Frequently asked questions

Can I write my own scoring rules?

Yes on Pro. Rule Builder lets you add + or − points on any signal or keyword match.

Does lead scoring work in every language?

Yes. Intent detection is multilingual across 40+ languages.

Where does enrichment data come from?

Public sources indexed by our partners — WHOIS, company websites, and publicly available firmographic data. No paid B2B database is required.