The difference between automatic and manual mileage tracking is not accuracy — a person with a notebook and an odometer can be perfectly accurate. The difference is coverage. Manual logs record the trips you remembered to record, and the gap between miles driven and miles logged is money: every forgotten business trip is a deduction that silently does not exist.
Why do manual mileage logs shrink?
Because remembering is the mechanism, and remembering fails in a predictable direction. Nobody accidentally logs a trip that did not happen; people constantly fail to log trips that did. The failure is worst exactly where gig work lives: many short trips, back to back, on days when you are busy. A single two-hour drive to a client is memorable. The fourth six-minute delivery of a lunch rush is not.
The usual patch — reconstructing the week on Sunday evening — replaces forgetting with estimating. Reconstructed logs have a recognisable texture: round distances, merged trips, and no record of which platform each drive served. As covered in the guide to audit-proof mileage logs, that texture is precisely what makes an auditor look closer. A log built from memory is weaker evidence and a smaller number at the same time.
What does an automatic tracker actually record?
Everything the phone can observe about the drive, at the moment it happens: when the vehicle started moving, the route it took, where it stopped, and the distance in between. Nothing has to be remembered at the kerb, which is where a manual log loses its miles — the trips you would never have written down by hand, the short ones, are captured with the same fidelity as the long ones.
Just as important is what a good tracker refuses to record. GPS positions are noisy; raw points drift through buildings and jump across junctions. A tracker that sums raw points overstates distance, and an overstated log is a liability, not an asset. The distance that gets saved should be the filtered, driven distance — a number you would be comfortable reading out to an inspector, with the route stored beside it as the evidence.
What is park-anchoring, and why does the endpoint matter?
Park-anchoring means the trip ends where the vehicle parked, not where the phone stopped moving. The distinction matters because delivery work continues after the driving does: you park, walk to a building, wait for a lift, hand over the order, walk back. A naive tracker keeps measuring through all of it and files the walk as driven distance.
The error is small per trip and systematic across thousands of trips, which is the worst kind of error a tax record can have — it inflates every entry in the same flattering direction. Walk-trimming closes the trip at the parking position and discards movement after it. The same mechanism solves the stop-at-the-door problem in reverse: pausing at a customer should not end a business trip and split one delivery run into fragments, so brief stops hold the trip open while a real park closes it.
Does background GPS tracking really drain your battery?
Far less than the myth says, because the myth is about how tracking worked years ago. Continuous, screen-on, navigation-grade GPS is expensive; modern background tracking is not built that way. Phones expose motion APIs that let an app sleep until movement that looks like driving begins, sample location at driving-appropriate intervals, and go back to sleep when the vehicle parks. The result is a tracker that is idle for most of the day by design.
The real battery problem runs in the other direction: some Android phones aggressively kill background apps to save power, and a killed tracker records nothing — the failure is silent and costs you a day of trips at a time. That is a settings problem, not a physics problem. It is fixed by exempting the tracker from battery optimisation, and a well-built app detects those settings itself and walks you through correcting them before they cost you a week of records.
Is manual entry ever the right tool?
Yes — as the correction and the exception, not the system. The phone left on the kitchen counter, the drive in someone else’s vehicle, the trip a dead battery swallowed: these need a manual entry, and a tracking app should make adding one easy rather than pretending it never happens. Manual entry is also where imported history lives, if you kept a spreadsheet before switching.
The point is the ratio. When automatic capture is the default, manual entry handles a handful of trips a year, made while the details are still fresh. When manual entry is the system, it handles every trip, and the forgetting problem does the pruning.
So which should you use?
Automatic capture with manual classification. Let the phone do the part it is strictly better at — being present for every drive and writing down what happened — and keep for yourself the part it cannot do, which is knowing whether a drive was business. A queue of recorded trips that you sort with a swipe at the end of a shift takes about two minutes; the end-of-shift routine post on this blog turns that into a habit.