What
Fleet managers tend to fall into one of two reflexes when ambulance tracking misbehaves. Either every dropped fix gets blamed on a setting that needs tweaking, or the whole device gets pulled for replacement at the first sign of trouble. Applied to the wrong failure type, both reflexes waste money. The useful question is not how bad the problem looks, but when and where it happens, because the timing of a lost fix points almost directly at whether a configuration change can help or whether you are looking at a physical design limit. Sorting failures into two operational windows makes that decision far simpler.The first window is tunnel transit. A device that goes quiet under an underpass, in a concrete parking structure, or beneath heavy steel is responding to a physical obstruction, not a fault. Most units behave this way, so the question becomes how much lag you can tolerate. Average delay in a commercial system runs somewhere between 2 and 10 seconds, though it can stretch past 30 seconds in dense urban areas. Those seconds matter: a 10-second delay in a tunnel can turn an 8-minute response into a 12-minute one. Since the cause is intermittent and tied to geography, it belongs in the software-first branch. Start by tightening the update interval from 30 seconds down to 10 seconds. That single change often shrinks the visible lag enough for dispatch purposes. If latency persists, reconfigure antenna placement, since a poorly seated antenna amplifies the same obstruction. If that still fails, a dual-network device is the logical next step, giving the unit a second path when one carrier fades. Each of these steps is a tuning exercise, and none of them requires retiring hardware.The second window behaves differently. Look for a stretch of silence beginning right after a hard stop, when the engine is off but the ignition is in accessory mode, and lasting up to 15 minutes while the crew is on scene. On the map this reads as a flat spot: the vehicle sits motionless while the stretcher is already moving. That pattern is rarely a sensitivity problem. The most common failure point is the power management circuit, which cuts the radio to conserve battery during the engine-off period. No dropdown menu corrects a device that intentionally stops transmitting. You can adjust intervals all day, and the unit will still go silent during the one phase of a call that matters most. Attempting to fix this through settings is the classic mistake that burns a budget cycle without resolving anything.Before you commit to replacement, though, rule out the network. A weak cellular signal paired with a stale fix can push the device back to last known cell tower triangulation, placing the unit several hundred meters from its real position, and in urban canyons a stale fix can sit as far as 500 meters off. That kind of error mimics a device defect. Check whether other vehicles on the same carrier show loss in the same zone at the same time. A shared outage points to the network, not your hardware.The decision boundary, then, is straightforward. Software tuning is justified when the failure is intermittent and tied to specific geography. Hardware replacement, or at minimum a redesign of the power management profile, is justified when the failure is consistent, tied to the engine-off window, and already cleared of any network cause. That is the difference between a fix that holds and a bill that returns.Building the budget case means documenting both patterns per unit rather than requesting a blanket refresh. Track how often each rig loses location in tunnels, how long each silence lasts after a hard stop, and whether neighboring units fail in the same spots. Raw signal health reporting from fleet management software gives you that evidence layer and separates device-level faults from network-level dead zones. When the request is backed by observed behavior on specific vehicles, the conversation shifts from a general upgrade to a targeted repair, and the spend lands where it actually changes response time compliance, geofence alerts, and the location data accuracy your billing depends on.