Definition
The degree to which manager behavior changes toward employees flagged as flight risks, measured as the difference-in-differences between flagged and unflagged cohorts before and after flag exposure.
The Measurement Gap
When a manager learns that an employee has been flagged as a flight risk, the manager's behavior toward that employee can change in ways that are invisible to every standard HR metric. Some managers invest more attention: scheduling stay conversations, offering development opportunities, advocating for compensation adjustments. Others disinvest: reducing project assignments, excluding the employee from stretch opportunities, beginning to develop replacement plans, or subtly withdrawing sponsorship. The disinvestment pattern creates a self-fulfilling prophecy: the prediction that the employee will leave triggers the very conditions that make leaving rational.
No legacy metric detects this behavioral shift. Manager one-on-one frequency might capture meeting cadence but not meeting quality or intent. Project assignment tracking measures work allocation but not whether changes in allocation were driven by the flight-risk flag. Performance ratings are annual snapshots that cannot isolate the moment when manager behavior changed. The employee's eventual departure is attributed to "voluntary attrition" and tallied in the turnover rate, with no trace of the prediction-triggered behavioral chain that accelerated it.
This is the deepest side effect of predictive talent models, not model inaccuracy, but model-induced behavioral change that corrupts the very outcomes the model was trained to predict. An organization deploying flight-risk scores without measuring FRBIS cannot distinguish between predictions that detected pre-existing departure risk and predictions that created departure risk.
Decision Supported:
Are flight-risk labels causing managers to change their behavior toward flagged employees in ways that accelerate rather than prevent departure?
Formula
Plain language: Compare behavioral indicators (project assignments, development opportunities, meeting frequency, rating trajectory) for flagged employees before and after flag exposure. Apply the same comparison to a matched unflagged cohort. FRBIS is the difference-in-differences, measuring the behavioral change attributable to the flag itself.
Notation:
FRBIS_d = (Behavior_flagged_post − Behavior_flagged_pre) − (Behavior_unflagged_post − Behavior_unflagged_pre)
Composite FRBIS = (1/D) × Σ FRBIS_d across D behavioral dimensions
Variables:
| Variable | Definition | Unit |
|---|---|---|
| Behavior_flagged_post | Behavioral indicator value for flagged employees in the period after flag exposure | Varies by dimension |
| Behavior_flagged_pre | Behavioral indicator value for flagged employees in the period before flag exposure | Varies by dimension |
| Behavior_unflagged_post | Same behavioral indicator for matched unflagged cohort in the same post period | Varies by dimension |
| Behavior_unflagged_pre | Same behavioral indicator for matched unflagged cohort in the same pre period | Varies by dimension |
| D | Number of behavioral dimensions measured (e.g., project assignments, development nominations, meeting frequency) | Count |
Proposed formula, pending validation pilots.