Feature, Freight News, IT Suppliers, Logistics, People

Execution Data a Goldmine for Fleets

[ May 19, 2026   //   Gary Burrows ]

By James Wee

(From FBJNA’s May Issue) Every fleet operation generates a massive volume of execution data every day: GPS pings, delivery timestamps, service durations, route deviations and more. This data holds significant potential to improve route planning and operational performance. In most organizations, however, much of it goes unused.

The data often goes relatively unexamined or sits in siloed systems, making it difficult to build structured visibility across the full operation and benchmark performance over time. Drivers may log excess miles from detours that aren’t consistently analyzed, while some routes finish early and others run late into the evening. The gap between what was planned and what happened often adds up to wasted fuel, unnecessary overtime and missed delivery windows.

The challenge is scale. Spanning hundreds or thousands of deliveries across routes with different customers, drivers, products and geographies, fleet execution data is too complex for manual analysis. No planner has the time or visibility to identify patterns across this volume. AI and machine learning, however, can quickly process this data, uncover key patterns, and surface specific insights that lead to measurable improvements. Yet despite this potential, only 5.6 percent of fleets are using AI broadly across their operations (https://tinyurl.com/4mfj8nr8).

Most systems were designed to serve specific functions, and the data they generate is rarely connected in a way that tells a complete operational story. Closing that gap is less of a technical challenge and more about connecting the data these systems are already generating, so AI can actually work with it.

Experienced planners and dispatchers may be skeptical of change and of technology claiming to reveal something they don’t already know. Compounding this is the perception that AI is meant to replace human judgment and oversight. In reality, it complements it by surfacing patterns that are impossible to see at the scale of daily operations.

Fleets that do start using AI to analyze their execution data are often surprised by the outcomes they can achieve. One compelling example is with service time estimates. Many operations assign a flat average to each stop, say, 10 minutes per delivery. But real-world service times vary enormously depending on numerous variables, such as industry, customer, product, vehicle type, traffic patterns, delivery locations and even geography. A five-minute discrepancy per stop may seem minor. More than 20 stops, however, that’s nearly two hours of lost productivity in a single day – and, over a month, the equivalent of several lost working days on just one route.

Machine learning addresses this by generating unique service time predictions for each stop based on historical execution data, learning from actual durations across variables like delivery volume, customer behavior, driver patterns, location characteristics and more. When planning precision improves, excess buffer time shrinks, idle capacity drops, and fleets can schedule more stops per driver within the same working hours. Early deployments of AI-powered route optimization software have found it is possible to increase route density by 30 percent (https://tinyurl.com/6tv66sr8)43 without adding vehicles or drivers.

Another reason many operations leaders hesitate to implement AI to analyze their fleet data is the assumption that it requires a major systems overhaul. But many of today’s AI and machine learning tools are built to work with the planning and mobile data that most fleets are already collecting.

It’s also not necessary to be a data analyst to get value out of these capabilities. The newest generation of AI tools also enables planners and dispatchers to interact with operational data using natural language – asking direct questions like why routes ran long, where delivery windows are missed, or where overtime is concentrated without building reports or exporting spreadsheets. Once that kind of visibility is in place, the planner’s job shifts from reacting to daily exceptions to identifying systemic patterns and addressing root causes.

It can be hard to know where to start when there’s so much data to look at. Service time accuracy is one of the most accessible starting points. It’s where the gap between planned and actual performance tends to be widest, and where data-driven corrections deliver measurable results fastest.

None of this happens overnight. Getting an organization to trust AI analysis over instinct, and to invest time in technology when there are trucks to load and routes to run, can be a genuine uphill effort.

But once fleets start the process, they can begin to see how it pays off and keeps getting better. Every improvement to service time accuracy makes the next round of planning tighter. And as AI and machine learning capabilities continue to advance, the returns on putting execution data to work will continue to grow.

The data already exists. The opportunity lies in connecting it, interpreting it and putting it to work – so fleets can move more efficiently from reacting to daily variability to continuously improving performance.

James Wee is general manager of fleet management at Descartes.

Tags: Descartes