Carrier-Lane Reliability Scoring: Gaming-Resistant Intelligence at Network Scale

The carrier scorecard problem

Every shipper runs a carrier scorecard. Most do it quarterly. A team pulls TMS data, calculates on-time percentages by carrier, formats it into a spreadsheet, and presents it at the QBR. These scorecards have three problems. First, they are retrospective: by the time the scorecard is built, the performance is weeks old. Second, they are single-company: the scorecard only reflects how the carrier performed for that one shipper. Third, they are gameable: carriers know which metrics shippers track and optimize behavior to inflate scores.

What we built

The Carrier-Lane Reliability Scoring API is a Loft action that returns a gaming-resistant, shipper-scoped On-Time Rate (OTR) score for any carrier on any lane. Continuously updated, network-informed, and designed to resist scoring inflation.

Shipper-scoped, network-informed

The score is scoped to the requesting shipper's data (preserving privacy), but the methodology is informed by network patterns. If a carrier has a systematic delay pattern on a lane across many shippers, that pattern influences the classification even if one shipper's sample size is too small to detect it alone. A single shipper with 50 loads per year on a lane cannot build a statistically meaningful reliability score. The FourKites network, with thousands of loads across dozens of shippers, can.

Gaming resistance

The API uses actual arrival timestamps from GPS, geofence, and yard gate events rather than carrier-reported times. It evaluates performance against original committed delivery windows, not renegotiated ones. It applies statistical methods to detect performance patterns consistent with gaming (suspiciously uniform on-time rates that do not match natural lane variance). And it factors severity of late deliveries, not just frequency: a carrier 15 minutes late on 10% of loads scores differently than one 6 hours late on 5%.

Continuous, not quarterly

The API updates as shipments complete. No batch processing, no quarterly refresh. When Tracy queries reliability before deciding how aggressively to follow up with a carrier, the score reflects the most recent performance. A carrier with a two-week driver shortage in a region shows degraded scores within days, not at the next quarterly review.

Why this matters

For agents, reliability scoring is the bridge between visibility and execution. Tracy intervenes earlier on low-reliability lanes. For procurement teams, it provides evidence for data-backed contract negotiations: network-validated reliability by lane, mode, and season instead of carrier self-reported metrics.