The photo that forgot where it was taken

A loan application from a cheap phone, on data that costs money. Two good ideas went into it.

The first: compress photos on the phone, from several megabytes to a few hundred kilobytes, so applying costs cents.

The second: read the location and capture time from the photo itself, on the server, so nobody can claim a photo of somewhere else.

Compression throws away the metadata. Each idea was right, each was documented, and together they would have left the fraud check reading nothing the moment the real upload was connected.

The design documents already called the check deterrence and triage, not proof. That honesty was the most reliable part of the system.