What
Start with the event record itself, because almost everything you need is already sitting in that one row. Open the start/stop location report and read the ignition-on and ignition-off times and places for that stop, the dwell time, and the engine state. Dwell time is simply the total minutes the vehicle sat stationary at that location between moves. Those few fields tell you whether the vehicle arrived somewhere, sat, and left, or whether the record looks scrambled from the beginning. If the ignition-on and ignition-off timestamps are out of order, or the locations jump a strange distance in a short window, you are probably looking at a data artifact rather than a real event. If the times and places line up cleanly, you are looking at something that actually happened, and now the question shifts to what kind of thing it was.The next fork is idle versus engine-off, and it matters more than most people expect. A running engine at a standstill is a different cost and a different conversation from a parked vehicle. If the engine stayed on for the whole stop, you are likely dealing with an idling problem, which points toward fuel burn, engine wear, and possibly a habits or waiting-time issue. If the engine was off, that same stop points somewhere else entirely: a delivery that ran long, an unscheduled break, a mechanical problem, or a driver waiting on a customer. Neither reading tells you the cause on its own, but each one narrows the field before you pick up the phone. Do not skip this check, because escalating an idling stop as if it were a breakdown wastes the driver's time, and treating a breakdown as idling wastes yours.Once you know the engine state, check the place against the boundary you already have. A geofence is a virtual boundary drawn on a map around a site such as a depot, customer dock, or fuel bay, and your platform usually stores entry and exit events alongside each stop. If the stop falls cleanly inside a customer dock boundary, the plan may simply have changed and nobody told dispatch. If it falls clearly outside every boundary, either the stop is genuine and unexpected, or your boundary is drawn wrong. Pay attention to stops that land right on the boundary line, because those are the ones that flip inside and outside on different days without anything changing on the ground. Before you redraw anything, confirm the site outline matches the physical site, then look at the threshold.The stop-detection threshold deserves a hard look whenever a stop sits in an unusual place. A threshold tuned for highway rest stops will misread a tight urban dock, and the reverse is true as well. If your settings expect long stationary periods before logging a stop, a quick in-and-out at a cramped yard can be split or merged oddly. That does not mean the vehicle did nothing; it means the record compressed or stretched what happened. Comparing the same location at the same hour across a handful of different days is the fastest way to separate a setting problem from a driving pattern. One bad afternoon proves very little. Two or three consistent appearances at the same spot start to look like something real.Position accuracy is the last thing to clear before you act on anyone. Ask yourself whether GPS tracking near that specific point is tight enough to trust at the distance you care about, which for a depot gate or a fuel bay can be a matter of metres. If accuracy readings near that spot are loose at the same hour on multiple days, the odd stop may be the position layer drifting rather than the driver doing anything unusual. Only after the engine state, the boundary, the threshold, and the accuracy check all point the same direction should you treat the stop as genuine.What each outcome licenses you to do is fairly narrow, and that is the point. A clean record with a running engine at a non-customer site earns a conversation about idling. A clean record with the engine off inside a customer geofence earns a question about the schedule or the loading window, not a warning. A stop that only appears near a boundary, or only under loose accuracy, earns an adjustment to thresholds or settings before it earns a coaching note. And whatever you find, remember that a start and stop location report is a reconstruction, not a verdict. Teams that connect these findings to broader reporting, or that watch the same events live through realtime vehicle tracking, tend to spot the difference between a pattern and a glitch faster because they are seeing the raw movement alongside the summary. One unexplained stop is worth investigating. It is rarely worth concluding from.