After 30 days of agent deployment, compare predicted vs actual outcomes to refine the rubric. Common adjustments: tighten criteria that didn't predict close, expand criteria that did, add disqualifiers based on losses.
What to gather
30 days of agent activity: which accounts were engaged, which replied, which booked meetings, which converted to opportunities, which lost. Tag each with their ICP score at time of engagement (turgo logs this automatically).
The four-quadrant analysis
- High score + converted (good predictions, keep doing).
- High score + didn't convert (false positives — what criterion needs tightening?).
- Low score + converted (false negatives — what criterion needs loosening or adding?).
- Low score + didn't convert (good predictions, ignore).
Common patterns to look for
- Industry criterion overrepresenting an industry that converts low — reduce weight or add a discriminating sub-criterion.
- Size band capturing too many sub-scale accounts that close at terrible terms — tighten the lower bound.
- Technographic criterion driving high scores but not closes — the technographic isn't actually predictive in your motion.
How to update the rubric
ICP → [rubric] → Edit. Change criteria, weights, or thresholds. Save creates a new version (the old version is preserved for comparison). The new version applies to future scoring; existing scores recalculate within minutes.
Don't over-tune
Statistical noise in small samples can mislead. If you've only had 5 closed-won deals in 30 days, don't make major rubric changes based on patterns in 5 deals. Wait for 20+ outcomes before significant changes; smaller adjustments are okay sooner.
