What a Photograph Proves and What It Does Not

Why this matters

Shops that got serious about photos usually got serious about volume, and volume is not the variable. A tech who shoots forty frames on a job and loses the dispute anyway has not been careless; he has been collecting the wrong kind of proof. A photograph is very strong at one narrow thing and structurally incapable of several others, and the ones it cannot carry are exactly the ones a dispute turns on: who caused it, in what order, and whether anyone agreed to it. Knowing the boundary changes what you shoot, and it usually means shooting fewer frames with more in each one.

The narrow thing a photograph is excellent at

A photograph is evidence of the state of a thing, at a place, at a time, to the resolution of the image. When all four of those are established, it is close to unarguable, and no written note competes with it.

That is a demanding sentence, because three of those four are not in the picture by default. The state is. The place is only there if something identifying the place is in the frame. The time is only there if the image carries reliable metadata or something in the frame fixes it. The resolution has to actually support the claim, which a distant shot of a small feature does not.

Most of the photos in a typical job folder establish exactly one of the four.

What sits outside the frame no matter how many you take

• Causation. A photo of damage proves damage. It has never once proved what caused it. Two photos, before and after, prove two states and the fact that they differ; the reason they differ is an inference somebody has to argue for.
• Sequence. A stranger looking at a set of images cannot order them. Image metadata can, and often does, until the file is sent through a messaging platform or social app, many of which strip metadata and recompress the image. The set that arrives at the other end may carry no reliable timestamp at all. Order has to be carried by a written record, not assumed from a folder.
• Authorization. A photo of a customer signing a tablet proves a person made a mark. It says nothing about what they were agreeing to unless the text above the line is legible in the same frame, which it almost never is.
• Absence. "There was no crack" is a much harder claim than "there was a crack." A photo can only support absence if the frame actually covers the whole area in question at a resolution that would have shown the feature. A wide shot from across a room does not support "the panel was undamaged."
• Scale. Corrosion in a close-up with nothing familiar beside it could be a fingernail or a forearm across. A known object in the frame fixes it. Without one, the reader supplies their own scale and it will match whatever they already believe.
• Condition over an interval. One photo of a gauge is a reading, not a hold. Proving a pressure held for a period takes two images of the same gauge with something in frame fixing the interval, plus the note that says what the test was.

Why the "before" shot is the weakest one you take, unless

Before photos are the ones shops are proudest of and the ones that fail most often, because they are usually close-ups of the thing about to be worked on. That frame establishes state and nothing else. The dispute, when it comes, is almost never about the state of the component. It is about something nearby that the tech never photographed, because it was not what he came for.

The fix is one wide frame per work area, taken from the doorway or the standing position, before anything is touched, showing the surroundings: floor, adjacent surfaces, ceiling below, whatever is stored nearby. That frame is the one that answers the accusation nobody has made yet. It costs about ten seconds and it is the highest-value image in most job folders.

Forty-one photos, and the two that decided it

A shop replaces a section of a pressurized line in a first-floor ceiling of a small commercial space. Nine weeks later the tenant reports staining on the ceiling in the room below the work, and the property owner asks the shop to cover the repair. The shop pulls its job folder: 41 photos.

The breakdown is typical. 33 of the 41 are component close-ups, about 80% of the set, showing fittings, the removed section, the new section, tooling in place. Each one proves a part existed and was handled. None of them shows the room, and several were forwarded to the owner through a messaging platform and came back stripped of their metadata. 6 more show the finished connection from various angles, which establishes the state of the work at completion.

That leaves 2, about 5% of the set, and they are the ones that matter.

The first is a wide frame the tech took from the doorway before starting, showing the work area and the ceiling of the room below through an open access panel, dry and unstained, with his tool bag in shot. State, place, and enough context to fix the visit.

The second is a photograph of the gauge at the end of the pressure hold, and by itself it is only a reading. What makes it work is the written note beside it in the job record: held at test pressure for 20 minutes, no observable drop, access panel left open and ceiling below dry at the time of leaving. The photo anchors the note; the note makes the claim. Neither one does the job alone.

The stain turned out to trace to a roof penetration on the other side of a joist bay, found when the owner's roofer opened it up. The shop's position was never that its photos proved it had not caused the stain, because no photo could prove that. Its position was that the area under the work was documented dry at the end of the visit and the line had held under test, so any theory that the shop caused it has to explain a nine-week gap with no reported water. That is what evidence actually does: it does not prove your innocence, it forces the other account to explain something.

Run the counterfactual. Strip out those two frames and the shop has 39 photos of fittings and a memory of a pressure test. The conversation stops being about a roof penetration and becomes a negotiation about who pays for a ceiling. The 33 close-ups did not fail because they were bad photographs. They failed because they were 33 answers to a question nobody asked.

The pairing rule worth adopting

Every close-up gets a second frame from a step back that includes a fixed feature of the room, taken from the same position. Two frames, roughly five seconds, and the close-up inherits the place from its partner.

Add a scale object where size is part of the claim. A tape, a standard tool, a coin-sized common object, anything a reader recognizes. Where a measurement matters, photograph the instrument reading and write the number down too, because a photo of a display that is glared out is not a reading.

The one thing a clip carries that a stills set cannot

Sequence is the gap in the list above that video actually closes, and only under one condition: the clip has to be unbroken. A single continuous pan from the equipment, across the room, to the area in question establishes that those things existed in that relationship at one moment, which no set of separate frames can establish on its own. That is worth deliberately doing where the dispute is likely to be about what was next to what.

The condition matters, because a clip assembled from several takes proves nothing more than the stills would have, and a clip that starts after the interesting part is a frame with extra file size. Keep them short, narrate what you are looking at out loud as you pan, and stop the recording rather than editing it. Sound carries too, which cuts both ways: everything said in the room while the camera runs is part of the record.

Where the photo cannot be taken at all

Some frames are not worth the exposure and a few are forbidden outright.

Never open an energized enclosure to get a photograph. If the interior needs documenting, de-energize at the disconnect and prove dead with a tester checked on a known live source before and after the reading, following NFPA 70E-2021 section 120.5 for the proving sequence and 29 CFR 1910.333(b)(2) for the safe work practice on electric utilization equipment; if it cannot be de-energized, photograph the outside and note what you could not see and why.

In a space with a gas odor, no phone is used inside at all, no switches or lights are touched, everyone leaves the area immediately, and the utility is called from outside. The photograph, if there is one, is taken after that and from outside. This is worth saying plainly because an article telling techs to document more is an article that will send someone into a space with a phone in their hand.

Do not reach a camera into rotating machinery or past a guard. A photo through the guard is worth exactly as much as one taken with the guard removed and costs nothing.

How to audit your own photo set

Take one finished job and try to answer these from the photos alone, without the tech in the room.

• Where was this taken? If more than half the frames cannot be placed in a specific room by their own content, you are shooting close-ups only.
• When was it taken? Check whether the copies you actually keep and send still carry metadata after passing through whatever channel you use. Assume the customer's copy does not.
• What did the area look like before anyone touched it? If there is no wide pre-work frame, the folder cannot answer the accusation that gets made.
• Which single image would you put in front of someone who disagrees with you? If the answer is "all of them together," you have volume standing in for evidence. Where the customer supplies the images rather than the shop, a sibling article covers how much weight to give them.

References

• NFPA 70E-2021, section 120.5, for the live-dead-live proving sequence, which binds through your employer's electrical safety program and through 29 CFR 1910.333(b)(2) for work on electric utilization equipment
• See related: The Photo That Protects vs the One That Incriminates, The Existing Conditions Photo Record, How to Use a Customer's Photo or Video as Real Evidence, The Record That Survives a Year of Memory