Search the visual record in plain language.
“Waterproofing photos, zone 4, before March.” Which is the question that gets asked in a dispute, and the one an archive organised by upload date cannot answer at all.
Zepth Core · Site Operations
Photos
A photo in a camera roll is a memory. A photo tagged to an activity is a record.
Last updated
Zepth Core module
Photos
AI agent built into the module
Capture inside every workflowAutomatic metadataPre-cover checklistsFindable archive
Overview
Site photos are the cheapest contemporaneous evidence a project produces — thousands of them, for free, as a by-product of people doing their jobs.
And most of that value is thrown away, not because the photos are bad but because they are unfindable and unattributable. Captured with a timestamp, a location, an author and a link to the record they evidence, a photo is proof. Captured in a camera app, it is a memory.
The metadata doctrine
What makes a photograph evidence is not the image. It is everything around the image: when it was taken, where, by whom, and what record it belongs to — which pour, which inspection, which non-conformance.
And that metadata is fragile in a very specific way. It survives capture and it dies in transit. Screenshot a photo and the metadata is gone. Send it through a messaging app and it is stripped and recompressed. Forward it twice and what arrives is an image of unknown provenance that could have been taken anywhere, by anyone, at any time — and the other side’s expert will say exactly that, because it is true.
So the capture point matters more than the camera. A photo taken from inside an inspection form is already attached to the inspection, already timestamped, already attributed. A photo taken in the camera app and uploaded later is a file that somebody will have to vouch for from memory.
Photograph what is about to disappear
The highest-value photographs on any project are of work that is about to be concealed. Waterproofing before the screed goes down. Reinforcement before the pour. Services before the ceiling closes. Fire-stopping before the wall is boarded.
These are worth more than every progress photo combined, and the reason is arithmetic: once the work is covered, the only way to see it again is to destroy something. A pre-cover photograph turns a five-figure opening-up exercise, three years later in a defects dispute, into a two-minute search. It is the single highest return on effort available anywhere in site documentation, and it costs thirty seconds.
The discipline is to make the shot a gate rather than a good habit — a checklist item that must be satisfied before the cover-up inspection is signed. Good habits do not survive a wet Friday with a concrete truck waiting.
Filing beats volume, and it is not close
Forty thousand unsorted images are barely better than none. Worse, in one respect: they create the belief that the project is documented, which survives right up until somebody has to find something.
Photos belong attached to the records they evidence — the inspection, the delivery, the snag, the diary entry, the NCR — rather than in a folder tree organised by the date somebody uploaded them. The test is simple and unforgiving: can you produce every photograph of a specific location, in a specific window, in under a minute? If not, the archive is a liability rather than an asset, because the other side will ask for exactly that and the answer will be a fortnight of scrolling.
And for the narrative, shoot consistent viewpoints on a cadence — the same positions, weekly or monthly. Tribunals and lenders both respond to a time-lapse in a way they do not respond to a thousand individually reasonable images, because a sequence from a fixed point is very hard to argue with.
The value
Why it matters
Photos arrive as evidence — timestamped, located, attributed, and attached to the record they prove.
Concealed work is documented before it disappears, which turns a future opening-up exercise into a search.
The archive is findable, so a disclosure request is a query rather than a fortnight.
The progress narrative exists as a fixed-viewpoint sequence, which is the form that persuades.
Capabilities
What you can do
01
Capture inside every workflow
From the inspection, the NCR, the delivery, the diary — so the photo is linked to its record at the moment it exists.
02
Automatic metadata
Time, location, author, and the linked record. Captured at source, because it cannot be reconstructed later with a straight face.
03
Pre-cover checklists
The shot that must exist before the cover-up inspection can be signed — a gate rather than a good habit.
04
Findable archive
Searchable by location, date, activity and record. The test is whether you can produce every photo of one zone in a window, in under a minute.
The workflow
How it actually runs
- 1
Capture in context, not in the camera app
From the inspection, the NCR, the diary entry. The photo arrives already attached to the record it proves, which is the whole difference between evidence and an image.
- 2
Auto-tag time, location and author
At the moment of capture — because this is the metadata that dies in transit, and it cannot be added back honestly afterwards.
- 3
Gate the pre-cover shot
A checklist item on the cover-up inspection, not a good intention. Good intentions do not survive a concrete truck.
- 4
Shoot the progress sets from fixed viewpoints
Same positions, on a cadence. A sequence from a fixed point is the narrative; a thousand reasonable images are not.
- 5
Retain against the limitation period
Not against the project. The claim arrives years after handover, and the photographs have to still be there — which is a decision somebody makes once, deliberately, or never makes at all.
AI that does the work
How AI changes Photos management.
Coverage-gap detection.
Zones with no recent photographs, and cover-up inspections signed with no pre-cover shot against them. The absence of a photograph is invisible by nature — unless something is looking for it.
Pre-cover prompting.
The checklist that fires before the concealment, listing the shots that will matter in three years to people who are not yet involved.
The engineer’s judgment stays in charge; the AI removes the latency and the blind spots.
Best practices
- Capture from inside the record, never from the camera app. A photo that arrives by message has been stripped of the metadata that made it evidence, and no amount of good faith puts it back.
- Gate the pre-cover photograph. It is the highest-return thirty seconds on the entire project, and it will not happen reliably unless something refuses to proceed without it.
- Retain against the limitation period, not the programme. The dispute arrives years after handover, and by then the decision to keep the photographs was made — or was not.
- Judge the archive by retrieval, not by size. Forty thousand unfindable images produce a feeling of being documented, which is worse than knowing you are not.
Dashboards & reporting
The photo record by location, activity, date and linked record — searchable, which is the only property that matters. Coverage gaps by zone. Cover-up inspections with no pre-cover photograph against them. Fixed-viewpoint progress sequences. And the disclosure view: every image of a given location in a given window, produced as a query rather than as a fortnight of scrolling.
Live dashboards
Drill-down & filters
Export to Excel / PDF
FAQ
Common questions
What makes a site photo usable as evidence?
The metadata, and the link. A date, a location, an author, and the record it belongs to — which pour, which inspection, which NCR. The image alone proves very little; it is the surrounding data that establishes it shows what you say it shows, when you say it did. And that metadata dies in transit: screenshot it, or send it through a messaging app, and it is gone.
Read the full answerWhat should be photographed before it is covered up?
Waterproofing before the screed. Reinforcement before the pour. Services before the ceiling closes. Fire-stopping before the boarding. Anything whose only other route to inspection is destroying something. These photographs are worth more than every progress shot combined, because once the work is concealed the alternative to a photograph is an opening-up exercise, three years later, in front of people who are already annoyed.
Read the full answerHow should site photos be organised?
Attached to the records they evidence — the inspection, the delivery, the snag, the diary entry — rather than in folders named after the day somebody uploaded them. The test is retrieval: can you produce every photograph of one location, in one window, in under a minute? Forty thousand unfindable images are worse than none, because they produce the belief that the project is documented.
Read the full answerAre phone photos acceptable in a dispute?
Photographs with intact metadata, captured in context and stored in a system with an audit trail, are ordinary and useful evidence. Photographs forwarded through a messaging app, screenshotted, or uploaded weeks later from a personal camera roll invite a question nobody wants to answer under cross-examination: how do we know when and where this was taken? The device is rarely the problem. The chain of custody is.
Read the full answerHow long should site photos be retained?
Against the limitation period, not the programme. A defects claim can arrive years after handover, and in some jurisdictions structural liability runs for a decade. The photographs either still exist at that point or they do not — and that is decided once, deliberately, by someone thinking about a dispute that has not happened, or it is decided by an IT clear-down nobody was consulted about.
Read the full answerSources
- SCL Delay and Disruption Protocol — contemporaneous records as the foundation of a defensible claim. Progress photography is the cheapest contemporaneous record a project generates.
- This page carries no stat band. The metadata doctrine and the pre-cover argument stand on their own, and we would rather say so than attach a number we cannot source.
Related answers
In the guides
Terms defined here
Zepth is the construction project delivery platform — it runs construction, procurement and asset management on one record, and does the work: reading the drawings, reviewing the submittals, matching the invoices and flagging the risks, with a human sign-off on anything consequential.
See it on your project.
A short, tailored walkthrough on your real workflow — no generic demo.