How Auto-Trip Detection Works

One of the trickiest problems in a mileage tracker is answering a deceptively simple question: "Is this person driving right now?"

Get it wrong one way and you're firing off notifications to someone sitting on their sofa watching TV. Get it wrong the other way and you're missing real trips, which is the entire point of the app. And all of this has to run on a phone that's trying to conserve battery, with iOS doing its best to kill background processes.

The basic approach

MileClear uses iOS's significant location change monitoring as the trigger. This is a low-power mode that wakes the app when the phone moves roughly 500 metres. The moment that happens, we check speed. If the phone is moving faster than about 15mph, we assume driving has started and begin recording coordinates at tighter intervals (100 metres).

Coordinates are buffered silently - no notification, no UI, nothing. We're just collecting GPS breadcrumbs in the background.

When does a trip end?

This is where it gets more nuanced. We don't end a trip the moment you stop - you might be at a red light, or queuing at a McDonald's drive-through, or briefly parked to drop off a package. Instead, we wait for five consecutive minutes of movement below 2.2mph (about walking pace).

Once that threshold is hit, we finalise the trip: calculate the distance, record the start and end times, save it to the local SQLite database, and mark it as "unclassified". It lands in your trip inbox waiting for you to tag it as business or personal.

The Driver/Passenger problem

Pure GPS detection can't tell the difference between you driving and you sitting in the back of an Uber. Both look identical from a location perspective: fast movement, then stopped.

The solution is a lock-screen notification. When MileClear detects a trip starting, it fires a notification with two action buttons: "Driver" and "Passenger". You can tap one without even unlocking your phone. If you're the passenger, the recording is cancelled and a 20-minute cooldown starts so you're not pestered again.

If you don't respond at all, the trip still gets recorded - we'd rather have a false positive you can delete than miss a real business journey.

GPS accuracy filtering

Raw GPS is noisy. A phone sitting still on a desk will drift by 10–15 metres. When you're doing speed calculations from sequential coordinates, that drift can make a stationary phone look like it's moving at 5mph - enough to cause false positives if you're not careful.

We filter out any location fix where iOS reports accuracy worse than 65 metres horizontal. We also require two consecutive readings above the speed threshold before registering movement. One blip doesn't start a trip.

Quiet hours

Nobody wants a phone notification at 2am because iOS decided to wake up the location service. Detection notifications are suppressed between 10pm and 7am. Trips still get recorded silently during those hours - the quiet hours only affect the Driver/Passenger prompt.

The battery reality

Background GPS is the number one complaint in every mileage app review section. MileClear uses significant location changes (not continuous GPS) when not actively recording, which is very low power. Even during active recording, 100-metre intervals are much less aggressive than the 50-metre intervals used during manual shift tracking.

The honest answer is: it does use some battery. Any app that tracks your location uses battery. But "significant location change" monitoring adds maybe 2–5% battery drain per day in practice - about the same as having Wi-Fi enabled. We're constantly tuning this.

What doesn't work yet

Train journeys are a known issue. A fast train exceeds the speed threshold and looks exactly like motorway driving from GPS. We're experimenting with CoreMotion activity recognition (the API that knows whether you're walking, cycling, or in a vehicle) to filter these out. It's not in the app yet, but it's coming.

If you spot a false positive, long-press the trip in your inbox and delete it. It takes two seconds and helps me understand where the thresholds need tuning.