# Predicted arrival and risk Every active shipment gets a predicted arrival time and a risk score: the chance it misses its appointment, with the reasons why. Here's how they're calculated today, what's coming, and their limits. ## How the predicted arrival works today The first version of the **Predicted ETA** adds up three things that anyone planning a trip would think about: | Input | What it adds | | --- | --- | | Routing | The remaining driving time on the road route from the truck's last position to the stop. | | Driver hours | Mandatory breaks and rest when the driver's remaining hours of service won't cover the rest of the trip. | | Facility dwell | The typical time spent at the remaining stops, based on that facility's history. | This approach doesn't need historical data to work, and every result can be explained. It's the baseline the platform launches with. > **In preview:** Predicted arrivals marked **Live data** come from the running platform. Risk scores are still part of the simulated data, and so is everything marked **Simulated data**. ## Reading a predicted arrival Open a shipment to see **Predicted arrival** for its final stop, and the ETA next to each remaining stop under **Stops**. The shipments list shows the final stop's ETA in **ETA (final stop)**. | Shown | What it means | | --- | --- | | **On time** | The ETA is before the end of the appointment window, with room to spare. | | **At risk** | The ETA is on time, but close to the end of the window (by default within 30 minutes). | | **Late** | The ETA is after the end of the appointment window. | | **No appointment** | The stop has no appointment window, so there is nothing to compare with. | | **Appointment** | How far the ETA is from the end of the window, for example **45 min after the appointment window**. | | **Change** | How much the ETA moved since the previous estimate. Small moves (under 5 minutes by default) show as **No significant change**. | | **Confidence** | How sure the estimate is, from 0 to 100%. It is lower when the distance had to be estimated without a road route. | | **Distance to go** | The remaining distance to the stop. | Times are shown in your own time zone when your profile has one, otherwise in the facility's time zone, always with the zone name. The thresholds behind **At risk** and **Late**, the time zone choice and the distance unit are settings, so your organization can change them. ## What shapes an ETA Under **What shapes this ETA**, each live ETA lists the things that added time or limited how precise it is, most important first: | Reason | What it means | | --- | --- | | **Driver break required** | The driver must stop for a break under the driving rules before reaching the stop. | | **Driver rest required** | The driver's daily driving time runs out, so a rest (for example 10 hours) is added. | | **Weekly hours run out** | The driver's weekly hours are used up, so a longer restart (for example 34 hours) is needed. | | **Truck driving ban** | Trucks can't drive in that country at that time (for example on Sundays), so the truck waits. | | **Waits for the appointment window** | The truck would arrive before the window opens and waits. | | **Planned time at stops** | Time planned at the stops before this one. | | **Distance estimated without a road route** | No road route was available, so the distance is a straight line times a road factor. Marked **Estimate**. | | **Based on the truck's recent speed** | The ETA uses the speed the truck has actually been driving. | | **Driver hours not reported** | The truck doesn't send the driver's hours, so the rules assume a rested driver since the last long stop. Marked **Assumption**. | > **Sample data is labelled:** Some calendars, such as truck driving bans, are still **sample data that hasn't been verified** against official sources. A reason that uses one is marked **Sample data, not verified**. Treat it as an indication, not a rule. Below the reasons you see the **driver-hours rules** used (for example US FMCSA or Canada federal), why those rules were chosen, and where the driver's hours came from. The model version and the time of the latest data are at the bottom of the card. ## Risk levels The **Risk of missing the appointment** is a percentage. It's grouped into three levels: | Level | Score | What to do | | --- | --- | --- | | **Low** | Below 30% | Nothing. It's on track. | | **Medium** | 30% to 59% | Keep an eye on it. It appears in **Exceptions**. | | **High** | 60% or more | Act now: contact the carrier, warn the receiver or move the appointment. | ## Reasons behind each score A score without an explanation isn't useful, so every shipment shows its reasons under **Why**. Examples: - **Driver hours: 1h 50m of drive time left** - **ETA is 65 min after the appointment** - **Traffic incident on I-20 near Birmingham** - **Snow forecast on Rogers Pass** - **Pickup dwell 2h 10m so far (typical: 1h 15m)** - **No location update for 3h 40m** ## Why the predicted arrival changes The prediction is recalculated each time something meaningful happens, not on a fixed timer: - A new position shows the truck is faster or slower than expected. - The truck stops for longer than a normal break. - Traffic or weather changes on the route. - The driver's remaining hours mean a rest stop is now needed. - A milestone is recorded, such as departure from pickup. Small changes are normal. A sudden jump usually means one of the reasons above, and it shows up in **Why**. ## What's planned | Feature | What it adds | Status | | --- | --- | --- | | Machine-learning ETA | Learns from past trips on each lane, carrier behaviour, time of day, facility dwell, traffic and weather, to correct the baseline. | Phase 2 | | Confidence ranges | An arrival window, for example "most likely 14:10, almost certainly before 15:00", instead of a single time. | Later | | Dwell prediction | Expected time at each pickup and delivery facility. | Phase 2 | | Anomaly detection | Unexpected stops, route deviations and trackers that go silent. | Phase 2 | | Scorecards | On-time performance and tracking compliance per carrier and facility. | Phase 2 | | AI assistant | Ask questions about your network in plain language. It only sees what you can see. | Phase 2 | > **PLANNED:** New models run in the background first and are compared against the current one. They only replace it when they're measurably more accurate. ## Limitations - A prediction is only as good as the latest position. If a shipment is stale, its ETA may be blank or out of date. - Events nobody reports (a breakdown, a closed dock) only show up once they affect the truck's movement. - Facility dwell is an average. A facility with little history gets a generic estimate. - Appointment windows are shown in the facility's local time zone. - Without a road route, distances are estimated and the confidence is lower. - When a truck doesn't report its driver's hours, the ETA assumes a rested driver, which can be optimistic.