The ML Learning System is your fleet's built-in intelligence engine. It watches every load, every trip, and every facility visit — then uses that history to predict what will happen next, so you can plan ahead and avoid surprises.
Think of it like an experienced dispatcher who remembers every delivery your fleet has ever made. Over time, that dispatcher learns which facilities always run late, which drivers are the most fuel-efficient, and which trucks are due for maintenance. The ML Learning System does exactly that — automatically, across your entire fleet, 24/7.
Every fleet that uses the system gets its own separate intelligence. Your data trains your predictions. Nothing is shared between companies.
Predicts how long your drivers will wait at each facility, based on historical patterns, time of day, and day of week.
Tracks fuel efficiency, safety scores, and driving behavior to spot trends and flag vehicles that may need attention.
Ranks drivers by fuel economy, safety, on-time delivery, and overall reliability — helping you reward your best people.
The ML Learning System follows a simple four-step cycle that repeats with every piece of data you add:
Every time a detention event is logged, a telematics reading comes in, or a trip is completed, the system captures it. Data can flow in from your ELD integrations (like Samsara or Motive), from GPS systems, or from manual entries your team makes in the app.
The system calculates averages, medians, and worst-case scenarios for each facility, driver, and vehicle. It also looks at time-based patterns — for example, a warehouse that's fast in the morning but always backed up after 2 PM.
When you need to plan a delivery, the system uses everything it has learned to estimate what will happen. For example: "Based on 50 previous visits to Walmart DC #4521, your driver will likely wait about 165 minutes if arriving at 2 PM on a Wednesday."
After the actual event happens, the system compares its prediction to reality. If it predicted 165 minutes but the actual wait was 168 minutes, that's 98% accurate. Over time, the system tracks its own accuracy and self-corrects. The more data it has, the more accurate it becomes.
Key takeaway: The system improves automatically. You don't need to configure, train, or tune anything. Just keep using the software normally and the predictions get better on their own.
The system makes four types of predictions, each designed to help you make better decisions:
Before sending a driver to a facility, you can see the estimated wait time. The system considers the specific facility's history, the expected arrival time (morning vs. afternoon), and the day of the week. For example, many distribution centers are slower on Mondays after a weekend of accumulated shipments.
By tracking engine performance, fuel efficiency trends, and idle patterns, the system can flag when a truck is likely to need maintenance soon — before an expensive breakdown happens on the road. You'll see which vehicles have declining health scores and approximately how many days until service is recommended.
The system learns what "normal" fuel consumption looks like for each vehicle and driver combination. When efficiency drops below expected levels, it can indicate mechanical issues, route problems, or driving habits that need attention.
Based on historical patterns, the system helps predict which drivers are the best match for specific loads — considering their fuel efficiency on similar routes, their on-time record at particular facilities, and their overall performance scores.
Every prediction comes with a confidence score — a number between 0 and 1 that tells you how much you should trust it.
You'll also see a P95 value for detention predictions — this is the worst-case scenario. If the P95 is 220 minutes, that means 95% of the time the actual wait will be shorter than 220 minutes. Use the P95 when you need to plan for the worst case.
Pro tip: New facilities will start with low confidence scores. After about 10-15 visits, the system usually has enough data to make reliable predictions. The more consistent you are about logging detention events, the faster confidence builds.