Definition
Variance in disruption recovery outcomes (wait time for rebooking, alternative service quality, refund processing) across fare classes and loyalty tiers, measuring whether the rebooking algorithm provides equitable service or systematically disadvantages lower-fare passengers.
The Measurement Gap
Why This Metric Didn't Exist Before
Before algorithmic rebooking, disruption recovery was handled by station staff making individual decisions. You couldn't measure systemic equity because there was no system: just hundreds of individual interactions. Staff might have biases, but those biases weren't encoded in rules that applied uniformly across the network.
Automated rebooking engines changed the game. They process hundreds of passengers simultaneously using priority rules that favor premium fare classes. This is arguably fair: premium passengers paid more and expect better service. But nobody has been measuring whether the disparity between premium and economy recovery is proportional to the fare differential or grotesquely disproportionate.
Here's the decision DREI supports: Is the rebooking engine creating a two-tier disruption experience where premium passengers are rebooked immediately while economy passengers wait hours? If business class pays twice the economy fare but recovers five times faster, the inequity exceeds the fare premium. That's not a service differentiation: it's an algorithmic equity failure that drives economy passengers to competing modes permanently.
Without DREI, you have anecdotal complaints from economy passengers and no data to distinguish legitimate service tiers from systemic disadvantage. The rebooking algorithm runs unopposed, optimizing for revenue recovery without any equity constraint. DREI provides the constraint.
Formula
Formula:
DREI = 1 − CV(Recovery_time_by_class)
Where CV = standard deviation / mean across fare classes
Variable Definitions:
| Variable | Definition | Unit | Source | Classification |
|---|---|---|---|---|
| Recovery_time_by_class | Mean time from disruption announcement to confirmed rebooking for each fare class | Minutes | Rebooking system | DERIVED |
| CV | Coefficient of variation across fare classes for the same disruption event | Dimensionless | Calculated | DERIVED |
Sign Convention: 1.0 = perfectly equitable recovery across classes. Lower = more inequitable. 0.0 = massive disparity.
Data Requirements:
| Data Element | Source System | Classification |
|---|---|---|
| Rebooking events with timestamps | Rebooking engine / ticketing system | NATIVE |
| Fare class of each affected passenger | Ticketing system | NATIVE |
| Recovery time per passenger | Rebooking system (disruption start to confirmation) | DERIVED |
| Alternative offered quality (same day, next day, bus, refund) | Rebooking system | NATIVE |
Edge Cases and Measurement Boundaries:
Premium passengers with more rebooking options present a legitimate confound: if first-class passengers can be rebooked on more alternatives because fewer first-class seats are needed, faster rebooking may reflect availability rather than algorithmic bias. Track whether the algorithm allocates economy seats to premium downgrade before offering them to economy passengers already waiting.
Small disruptions with fewer than 10 affected passengers per class have limited statistical meaning for DREI. Aggregate across multiple disruptions. Minimum sample: 5 disruption events affecting 20+ passengers across multiple fare classes per measurement period.
Proposed formula, pending validation pilots.