A photograph on its own is evidence of almost nothing. It can be old, it can be of a different property, and the data attached to it can be rewritten in seconds with free software. Here is how the chain is built instead — stage by stage.
You enter the property address. It is resolved to a coordinate and stored against the job before any link exists. Everything that follows is measured against this point, so the target is set before the person onsite is involved.
The person onsite gets a link by SMS or email. One property, one submission, one window of time. They never create an account, install anything, or see your dashboard. The link is the boundary — it cannot be forwarded and quietly reused against a different job.
The link opens the camera in the phone’s browser. When each photograph comes back to the page, a fresh location reading starts and is sent with the image. The recorded location does not come from the file. Metadata inside an image can be rewritten in seconds with free software, so a photo that carries its own location proves nothing.
Before a photo is stored it is resized and saved as a JPEG. A SHA-256 hash of that stored file is printed in the report. Anyone can recompute it later from the file in the download, without access to our systems. A match means the file is byte-for-byte the one stored. No match means it has been altered.
Every submission is checked for signals that it does not fit the story it tells: how far the capture point sits from the address, whether the connection is being deliberately anonymised, and whether the submission is internally consistent. An ordinary VPN is not treated as suspicious — a large share of iPhone users route traffic through Apple’s relay without ever choosing to, and flagging them would bury the signal that matters.
Photographs arrive with the location, the timing, the checks and the full activity trail. You decide whether to approve. Nothing is filed on your behalf, and no language model is involved in what appears on a filed report — every check is deterministic code, because a report has to mean the same thing at audit in five years.
On approval, the Property Verification Report is produced: photographs, a map of the capture point, the hashes, the checks and the timeline, in one PDF. The stored photos stay available as a ZIP. The hash on the report is the hash of those files, so it can be checked independently. Anyone can check a downloaded PDF at onsitedata.com.au/check-report — the file is hashed in their browser and never uploaded.
The mechanism matters less than the artefact it produces.
View a sample report →