The Predictive Impact Model replaces influence bubbles, which used Pearson bivariate correlation to estimate driver impact on engagement.
Key Rules
- Influence bubbles (legacy): Measured statistical correlation between an individual driver and the organization's overall Engagement Score. A large, filled bubble meant the driver was highly correlated with engagement. A small bubble meant lower correlation.
- Predictive Impact Score (current): Uses a decision tree-based regression algorithm trained on over 600,000 completed engagement surveys. Scores quantify the expected point change in engagement if responses to a specific statement improve.
- Influence bubbles operated at the driver level. Predictive Impact Scores operate at the individual statement level, then aggregate by team, department, or custom segment.
- The legacy model identified correlation. The current model identifies predicted causal impact — which statements, if improved, would move engagement scores.
Common Misunderstanding
A high correlation (large influence bubble) did not mean that improving the driver would raise engagement scores. Correlation measures relationship strength, not directional impact. The Predictive Impact Score is designed to identify where action produces measurable improvement.