The Predictive Impact Model is an AI-powered analysis tool built into 15Five engagement surveys. It calculates which Engagement Driver statements, if improved, would most increase overall engagement scores — for the organization as a whole or for specific groups. It is available in the Engage and Total Platform pricing packages and is embedded directly in engagement survey results: anyone with access to engagement reporting can view Predictive Impact Scores within their visibility settings.
For the science behind the model, read the Predictive Impact Model whitepaper.
Feature Structure
The model uses decision-tree regression to calculate how changes in individual statement responses would influence overall engagement scores.
- Predictive Impact Scores — Calculated for Engagement Driver statements only. Each score quantifies how much that Driver statement affects overall engagement; scores with larger negative values indicate higher-priority improvement areas. Engagement statements, Manager Effectiveness statements, Organizational Commitment statements, and custom statements do not receive Predictive Impact Scores.
- Group-Specific Analysis — Scores are calculated at the individual level and aggregated by team, department, or custom segment. Each group's highest-impact statements can differ from the organization's overall results.
- Where scores appear — Predictive Impact Scores appear in three places in engagement survey results: the Statements tab; the Driver tab (Compare Statements by group); and the Heatmap report when the Drivers of Engagement survey is selected.
Key Rules
- The model is embedded in engagement survey results. Anyone with access to engagement reporting can view Predictive Impact Scores for Engagement Driver statements, limited to their visibility settings. Access is not gated to any dashboard role.
- Scores are recalculated for each engagement survey campaign — results are not carried forward from prior campaigns.
- The model draws on 15Five's dataset of over 600,000 completed engagement surveys to identify patterns. It does not use only your organization's data in isolation.
- Group-level scores require sufficient response volume to calculate. Groups with low response counts may not display scores.
- The Predictive Impact Model replaces legacy influence bubbles. See How does the Predictive Impact Model differ from influence bubbles? for comparison.
What You Can Do
Analyze
- Analyze Engagement Results Using the Statements Report
- Understand the Engagement Report
- Filter and Segment Engagement Campaign Results
- Heatmap Report — Engagement Survey Results
- Create an Action Plan from Engagement Survey Results