Why on-time performance is the metric that matters most
A late pickup in NEMT is not an inconvenience — it can mean a shortened dialysis treatment, a missed procedure slot, or a rider stranded outside a clinic at closing time. Brokers track provider on-time performance and shift volume toward reliable providers, and many contracts set explicit OTP thresholds that vary by broker and state.
Late will-call returns are often the most damaging kind of lateness, because the rider is already tired and the facility is calling your dispatch line every ten minutes. Improving OTP improves nearly every relationship your business depends on.
Measure OTP properly before trying to fix it
You cannot improve a number you are estimating. Define on-time using your contract terms — brokers differ on whether the window is measured against the scheduled pickup time or the appointment time, and how many minutes of grace apply — and measure against GPS-verified arrival timestamps, not driver recollection.
Then segment the number. A fleet-wide 90% can hide a 70% problem on early-morning dialysis runs or one chronically overloaded driver.
- Use GPS or driver-app timestamps, not handwritten times
- Segment OTP by time of day, day of week, driver, vehicle, facility, and trip type
- Track A-leg (pickup) and B-leg (return) performance separately — returns usually score worse
Diagnose the real causes of lateness
Most lateness traces back to a handful of predictable causes rather than bad drivers. Dwell time is the big hidden one: the plan assumes five minutes at the door, but a rider in a wheelchair on the third floor of a building with a slow elevator takes fifteen, and every trip after that inherits the delay.
Other common culprits: drive-time estimates that ignore rush hour, schedules packed with zero slack so one hiccup cascades all day, will-call returns dispatched reactively with no vehicle nearby, and dispatchers finding out about delays only after the rider calls.
- Underestimated dwell time at pickups and drop-offs
- Static drive-time estimates that ignore traffic patterns
- Over-tight schedules with no recovery buffer between trips
- No live ETAs, so problems surface after the pickup is already missed
Fix the schedule: buffers and realistic dwell times
Build rider-specific dwell times into the schedule instead of a single default. The rider who needs door-through-door assistance and wheelchair securement is not a five-minute stop, and pretending otherwise makes the whole manifest fictional.
Add deliberate recovery buffers at points where delays historically accumulate — after hospital discharges, around school-hour traffic, before tight appointment clusters. A schedule with 10% slack that holds up beats a perfectly dense schedule that collapses by 9 a.m.
Stage vehicles for returns and peak windows
Will-call returns are late when the nearest vehicle is twenty minutes away at the moment the rider is ready. Instead of dispatching returns purely reactively, stage vehicles near facilities with predictable release patterns — dialysis centers discharge in waves tied to treatment start times, so position capacity accordingly.
Ask facilities for advance notice where possible; even a 15-minute heads-up that a rider is finishing changes a late return into an on-time one.
Use live tracking with early-warning alerts
The cheapest minute to save is the one before the pickup is missed. With live GPS tracking and predicted ETAs, dispatch software can flag a trip trending late while there is still time to swap drivers, call the facility, or re-sequence the route. Axen, for example, surfaces at-risk trips on the dispatch board so dispatchers act on exceptions instead of watching every dot on the map.
Pair alerts with a standard playbook: who calls the rider, when a trip gets reassigned, and what gets logged. An alert nobody acts on is just noise.
Communicate with facilities like partners
Facilities cause lateness too — riders not ready at pickup time, discharges that slip by an hour, front desks that do not know your driver arrived. Give your regular facilities a direct dispatch contact, share ETAs proactively, and agree on where drivers wait and how arrival is announced.
A monthly five-minute conversation with the charge nurse at your busiest dialysis center prevents more late trips than most software features.
Run a weekly OTP review loop
Improvement sticks when it is a routine, not a project. Once a week, pull the late trips, group them by cause, and pick one fix to test — a dwell-time adjustment, a staging change, a schedule buffer — then check the next week whether that segment improved.
Share the OTP number with drivers and dispatchers, and celebrate the trend rather than punishing individual late trips that had legitimate causes. Drivers who trust the measurement will tell you where the schedule is unrealistic, which is exactly the information you need.
Related resources
Frequently asked questions
What counts as on-time in NEMT?
It depends on the contract. Brokers and states define their own windows — commonly a set number of minutes around the scheduled pickup time, sometimes measured against the appointment time instead. Always measure OTP using the definition in each contract you serve.
What is a good on-time percentage for NEMT providers?
There is no universal benchmark, and contract thresholds vary by broker and state. The practical approach is to know your contractual minimums, baseline your actual GPS-verified performance, and improve steadily from there.
Why are return trips late more often than pickups?
Return trips, especially will-calls, have unpredictable start times because they depend on when the appointment or treatment actually ends. Providers improve them by staging vehicles near facilities with predictable discharge patterns and getting advance notice when riders are finishing.
Does route optimization improve on-time performance?
It helps when the underlying inputs are honest. Optimization builds schedules that respect time windows and realistic drive times, but if dwell times are underestimated or buffers are stripped out, even an optimized plan will run late.