SD-301g · Module 2

Quality Thresholds and Monitoring

3 min read

Quality degrades when volume increases. This is the fundamental tension of personalization at scale. The monitoring system catches the degradation before it reaches the prospect. Three quality metrics: accuracy rate (percentage of personalization claims that are factually correct), relevance rate (percentage of messages where the personalization connects to the prospect's actual situation), and reply rate by personalization tier (do insight-personalized messages actually produce more replies?). If accuracy drops below 95%, pause and investigate. If relevance drops below 80%, the research pipeline needs better filtering. If tier-3 messages produce the same reply rate as tier-1, the personalization is decorative, not functional.

Do This

  • Track accuracy, relevance, and reply rate by personalization tier weekly
  • Investigate every factual error — a wrong personalization claim is worse than no personalization
  • Establish minimum quality thresholds and pause volume when they are breached

Avoid This

  • Measure personalization success only by volume — sending more bad messages is not progress
  • Accept a 90% accuracy rate — the 10% of prospects who receive wrong information remember it
  • Skip quality monitoring because "the AI is getting better" — verify, do not assume