How scoring works
Every lead has a score from 0 to 100: fit (0–60) plus intent (0–40). The lead page explains every point.
Fit (who they are)
| Rule | Default weight |
|---|---|
| Industry matches your ICP | 14 |
| Employee count in your range (half points if unknown) | 8 |
| Country or city matches (a third of the points if unknown) | 12 |
| Uses a technology from your ICP | 8 |
| A keyword from your ICP appears in their profile | 8 |
| The person's title matches a persona | 10 |
The fit total is scaled to 60.
Intent (why now)
| Signal | Default points |
|---|---|
| Funding | 18 |
| Hiring | 14 |
| New leader | 14 |
| Post engagement | 14 |
| Custom signal | 12 |
| Website visit (+6 for hot pages such as /pricing) | 12 |
| Tech-stack change | 10 |
| Keyword mention | 8 |
| Company news | 8 |
| New Indian company (MCA) | 8 |
Signals lose half their points every 14 days (the half-life). The strongest signal of each type counts fully; each repeat of the same type adds a quarter of its points. Intent is capped at 40.
Changing weights
Open the agent, edit its ICP, and change any weight or the half-life. New and refreshed leads use the new weights.
Learning from replies
Once enough replies and meetings have come back (at least 20 outcomes, and 5 for a given rule), the agent page suggests weight changes with a reason for each, capped at ±30% per round. You accept or dismiss them; nothing changes on its own. See Learning loop.
Leads below the agent's minimum score are kept as below threshold. They don't count against your allowance, and they are promoted automatically if new signals push them over the minimum.