Ce guide est actuellement disponible en anglais et en chinois.

Workflows / JevModel

Score leads on a useful rubric

Make sales prioritization easier to test and revise.

Dernière mise à jour:

Write levels as behaviors

A useful Score rubric describes visible signals: no stated need, some fit, a requested demo, or active procurement. Avoid labels like “good” and “excellent” without evidence. Keep the state current and exclude data you do not have a reason to use.

Measure downstream value

Compare score bands with actual follow-up outcomes. If teams change their behavior based on a threshold, review the threshold against recent data rather than relying on an early sample.

Run and interpret a Score example

The linked GitHub lead-scoring directory includes a synthetic lead, four ordered rubric levels and an illustrative response. Run from the repository root. Set JEVMODEL_API_KEY and add --live only for one real request. Inspect data.answers.priority.score; it may be fractional and is not a conversion probability or inherently a 0–1 value. Check the returned scale, then evaluate score bands against labeled leads and follow-up outcomes. No CRM updates or emails are performed.

Batch text classification with the JevModel APIEvaluate a Jev workflowJevModel API: request and responseIntent Classification for Product & Support Text: Rules, LLMs, and Typed APIsSpam Detection API for Forms and Comments: Turnstile, Akismet, or a Jev Decision LayerAI People Search with Jev: Screen Candidates Against Your ICP Before Paying for Email Lookups
node quickstart/node.mjs lead-scoring
# Or: python quickstart/python.py lead-scoring

JevModel est indépendant et n’est pas affilié à TypeSafe AI.