Before: Days of Waiting in Vain
We sent 50 buyer emails every week, but had no idea when replies would arrive. We wasted time tracking emails that never got responses. The team could only answer 「Will this buyer likely reply?」 based on experience and intuition.
After: The Signal the Scorecard Speaks
We collected 3 months of email metadata (subject line length, body word count, send day, buyer industry) and actual response outcomes, then trained Claude Code to learn the patterns. Now every new email automatically receives a score.
Buyer Response Prediction Score System
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
80+ points: High response likelihood within 48 hours (intensive tracking)
50~79 points: Response within 1 week (regular check-in)
Below 50: Cool-down period (resend after 2 weeks)
What Actually Changed
1. Contact timing became precise
Emails sent on weekday mornings tend to score high, while Saturday sends score low. The team now concentrates efforts on Tuesday through Wednesday mornings.
2. Non-responsive buyers identified quickly
Buyers consistently scoring in the 40s signal zero interest in our products. This silent nudge tells us to stop chasing and move to fresh prospects.
3. Team psychology shifted
Instead of applying equal effort to every buyer, the team now wonders 「Why hasn't this 75-point customer responded yet?」 and starts experimenting with better email copy.
Surprising Discoveries
• Response patterns vary by industry. Some sectors show 80% reply rate on first contact, while others only respond to second emails.
• Buyer names containing certain regional markers boost response probability by 10 points.
• Overly long body text (300+ words) drops predicted scores by 15 points.
Next Phase
Google Sheets scoring automation works well, but still requires some manual calculation. We plan to use Claude Code to convert this into a Windows background task, so every morning at 6 AM, all yesterday's emails will be automatically scored.
The most striking thing is that the bot sees patterns before we do. Now we're learning to read its signals.