Lead Scoring Model Audit and Calibration Prompt

You are a marketing analytics lead auditing a lead scoring model. Analyze whether the current score is predictive, trusted by sales, and calibrated to actual conversion outcomes.

Business model: [B2B SaaS, marketplace, services, ecommerce, etc.]
Lead sources: [paid search, organic, webinars, outbound, partners, trials]
Current scoring rules/model: [demographic points, behavior points, fit score, ML model, intent data]
Fields available: [firmographics, engagement, product usage, campaign touches, CRM stages]
Conversion goal: [MQL to SQL, demo booked, opportunity created, closed won]
Time window: [last 90 days, 6 months, 12 months]
Known issues: [too many false positives, ignored by sales, source bias, stale scores, low volume]
Sales feedback: [qualitative notes]
Constraints: [CRM/marketing automation platform, data quality gaps, privacy limits]

Create:
1. Audit plan with required data tables and joins.
2. Diagnostic metrics: conversion by score band, precision/recall, lift, calibration, source mix, and time-to-convert.
3. Bias and leakage checks for fields that may overstate quality.
4. Segment analysis by lead source, company size, region, persona, and product interest.
5. Threshold recommendation for MQL/SQL handoff with tradeoffs.
6. Recalibration plan for scoring weights or model features.
7. Sales validation workflow and feedback loop.
8. Dashboard layout for ongoing monitoring.
9. Caveats for low-volume segments and seasonality.