The roof photos you can never take again

20 August 2026 · 8 min read

Most contractors do not lose supplements for taking too few photographs. They lose them for taking a hundred of the wrong subject.

Hail hits get photographed. Everyone photographs hail hits. The frames that decide the file are usually of things that were never damaged at all — the eave edge, the felt under the first course, the number of layers, the vents that were already up there. Those frames share one property that no amount of care later can restore: they can only be taken while the old roof is still on the building.

A shovel is a one-way door. At seven in the morning the roof is evidence. By noon it is in a dumpster, and everything you did not photograph has quietly turned into an assertion.

The frames that cannot be retaken

Take these before the first shingle comes off — not before the job, before the first shingle. Each one exists to defend a specific line item, and each one answers a specific rejection.

• Eave edge, shot from the side so the edge reads in profile, tape in frame. Defends drip edge, gutter apron and starter placement. Answers: it was not there before the loss.
• Rake edge from the gable end, same treatment. Defends rake length of drip edge, which is a separate quantity from the eaves. Answers: an estimate that carries eaves only.
• Underlayment at the eave, once the first course lifts. Defends underlayment type, and whether a membrane exists and where it stops. Answers: felt is felt, and no membrane is required here.
• Layer count: a cut-back square or the first cut, tape across the stack, shot from the cut edge. Defends additional layer tear-off. Answers: one layer was estimated.
• Existing ventilation from outside — every unit, and its type. Defends vent count and like-kind replacement. Answers: box vents would have been sufficient.
• The same ventilation from inside the attic, plus the soffit intake. Shows whether the exhaust was continuous, and whether intake was blocked or baffled. Answers a quantity dispute that otherwise has nothing on your side of it.
• Each valley, showing open metal or a closed cut. Defends valley metal and valley lining. Answers: a closed valley could be used instead.
• Wall and chimney flashing in place: step, counter, apron, kickout. Defends replacement rather than detach and reset, and any siding or masonry involvement. Answers: detach and reset is sufficient.
• Every penetration, one numbered frame each. Defends boot and jack quantity. Answers: only one allowed.
• Decking from inside the attic, before anything moves. Defends sheathing replacement, re-nail, board type and span. Answers: pre-existing condition.
• Hip and ridge cap in place, close enough to read the profile. Defends cap type and quantity. Answers: it is already in the shingle line.
• Rooftop equipment: mount, penetration, cable route. Defends detach and reset, itemized. Answers a lump sum with nothing behind it.

The attic frame is the one that gets skipped and the one that carries the most weight. The decking rejection to plan for is the one about time: the wood was already bad. The only thing that argues with a claim about time is a photograph taken earlier. From below, before the crew arrives, it takes a few minutes.

Where the code argument lands

Some of these frames are not about condition at all. They are the surface a code argument has to land on.

Drip edge at eaves and rakes sits at R905.2.8.5. Valley lining sits at R905.2.8.2. The ice barrier requirement sits at R905.1.2, and whether it reaches that particular address is decided by the climatic and geographic criteria at R301.2 as adopted locally. Metal panel roofs carry their own flashing requirement at R905.4.6. Adoption happens state by state and often by local amendment, so the accurate sentence is that your state adopted a particular edition of the IRC, not that the IRC as such says something — and the edition, along with the section number in that edition, is worth confirming with the authority having jurisdiction before you put it in writing.

None of that is proved by a photograph. What the photograph proves is the condition the requirement gets applied to: that there was no drip edge, that the valley was open metal, that the membrane stopped short of where the eave actually is. Without that frame the code paragraph has nothing to land on and reads as theory. There is a longer treatment of one of these fights in the drip edge write-up.

One frame, one line item

Here is the test. Point at any photograph in the file and ask: which line does this defend? If the answer takes more than a second, that photograph is decoration.

A stack of images at the back of a packet is not an argument. It is homework handed to a reader who did not walk the roof, and a reader rebuilding your scope out of a pile is rebuilding it from their reading of it rather than yours. Put the frame beside the item it supports, with a caption naming the slope and the measurement and nothing else.

Run it backwards too. Go down the items you are claiming and mark the ones with no frame. Those are the items you are asking someone to take on your word. The ones that turn on a condition nobody can see any more are the hardest to defend that way, because there is no longer anything to go back and look at.

Framing: a measurement, not an impression

Significant pitting on the north slope is an opinion, and an opinion invites a second opinion. North slope, a test square chalked out and measured, each mark circled and counted, the count written on the surface, shot from directly above and then close — that is a measurement. Someone else can check it. That is the entire difference, and it is most of the job.

• Scale that returns a number. A coin says roughly that big. A tape says a dimension. Lay the tape so the reading is legible inside the frame, not beside it.
• Wide, then medium, then close — one sequence, one subject. A close-up on its own is unlocatable. The wide frame says which plane this is, the medium frame carries extent, the close-up carries detail. Shot out of order they read as three unrelated pictures of a roof.
• Countability. Report hits as a count per slope, never as a description. Chalk the perimeter of the test area so the boundary is visible, chalk each mark, and get the count and the size of the square into the frame itself rather than only into your notes. If the slope carries more marks than the square, keep marking — a count that stops on a round number is hard to tell apart from a count that stopped early.
• One naming scheme for slopes. Take the compass directions off the measurement report and use those same names in the photo filenames, the captions and the narrative. A file that calls one plane north in one place and front in another creates a question that has nothing to do with the roof.
• An anchor frame per set. One shot with the street number visible ties everything after it to that property, which is one question nobody has to raise later.
• Do not process the frames. No filters, no crop that cuts the scale out, no re-export through an app that discards the capture date. If your workflow strips dates anyway, put a dated marker board in the first frame of the day and let the sequence carry the rest.

How many is enough, and when it turns into too many

There is no correct total. The number that matters is coverage per item, and per item it is small: usually one wide and one close, plus, wherever a quantity is claimed, the frame that lets somebody count.

Too many looks like this. A run of near-identical shots of the same mark. The same slope from several angles that all show the same thing. Photographs of items you are not claiming at all. Every extra frame is one more place for a reader to find something that looks like it contradicts a quantity, and it buries the frames that were doing work. Volume is not thoroughness. Coverage is.

The tear-off morning sequence

1. Before the first shovel. Everything in the list above, attic included.
2. At the first cut. Layer count, tape laid across the stack, shot from the cut edge so the layers separate visually. That one frame records something nothing else will be able to show an hour later.
3. Deck exposed. A wide of the plane, then each bad sheet with the tape in frame, then one overall shot with the bad sheets marked so the count reads in a single image.
4. At the moment of discovery. Anything that was not in the estimate gets photographed before it is touched, and reported the same day. The frame establishes the condition; when you sent notice establishes the timing. Those are two different questions, and an invoice answers neither.
5. Before dry-in. The last frame of the open deck. Once the membrane is down, everything underneath it is a statement about the past.

This is not the crew's trade and it slows their morning, so it does not happen unless someone owns it. Whoever owns the file owns the camera. Otherwise the first pass goes missing on exactly the jobs where it mattered.

What a photograph cannot do

A photograph establishes a physical fact: this material was there, in this condition, at this measurement, at this time. That is all it does, and it is a great deal.

It does not settle what is owed. A caption that argues coverage is doing the narrative's job, and it puts an opinion where the file needs an observation. Keep the caption to what is visible. Keep the reasoning in the narrative, item by item, where each rejection can be answered on its own terms — there is a line-by-line list of those rejections here.

The habit underneath all of it is simple, and it explains why the frames come before the argument. You are not photographing damage. You are photographing a building that is about to stop existing in its current form, so that later, when somebody asks what was under there, the answer is not a memory.

Written for contractors preparing their own documentation. This is not legal advice, it is not a reading of anyone's policy, and it does not tell you what any carrier owes on a given file. Code adoption, section numbering and local amendments vary by jurisdiction and by edition — verify the adopted edition and the section against the authority having jurisdiction for the property.