Guide
What a damage photograph can and cannot establish
A photo estimate is not a worse estimate. It is an estimate written from a narrower evidence base, and the useful question is which lines that evidence actually reaches. The corpus answers that panel by panel.
Last reviewed . Every figure below is read from the measured corpus at build time, so it moves when the corpus does.
Concealed components sent to replacement
99%
124 decisions across 4 components
Not one of these is visible in a photograph of an intact bumper cover.
The fender, the closest thing to a coin toss
72% replaced
61 decisions
The door shell is close behind at 63% of 93. On panels like these the documentation is doing the deciding, not the panel.
The camera sees the cover. Most of the estimate is behind it.
The components this corpus replaces most reliably are the ones sitting directly behind the part a photograph shows. An absorber, an impact bar, a wheelhouse liner and a radiator support are all inside the structure, and a bumper cover that took the hit can look survivable while every one of them is scrap. That is not an argument against photo estimating. It is the reason a photo estimate and a teardown estimate are two documents rather than one.
| Component | Replaced | Decisions |
|---|---|---|
| radiator/condenser | 100% | 55 |
| bumper impact bar | 96% | 25 |
| fender liner | 100% | 24 |
| bumper absorber | 100% | 20 |
The panels a photograph settles
These went to replacement every single time they appeared in the corpus. Mostly bolt-on, mostly not repairable in any meaningful sense, and a photograph showing the damage is enough to write the line. The sample sizes vary by an order of magnitude across this list and they are printed for that reason: the ones at the bottom are at the published minimum of 6 observations, and 6 identical outcomes is a hint that a pattern exists rather than proof of one.
| Panel | Decisions |
|---|---|
| emblem/nameplate | 59 |
| radiator/condenser | 55 |
| fender liner | 24 |
| bumper absorber | 20 |
The panels a photograph only argues
These went both ways. The closer a panel sits to the middle of this table, the more the outcome depended on the estimate rather than the damage, and that is exactly where the photograph is doing real work. A fender creased at its rear edge is a fender job. A fender creased at its rear edge and sitting hard against the door is a fender, a door and an alignment problem, and only the wider frame shows the second one.
| Panel | Replaced | Repaired | Decisions |
|---|---|---|---|
| rocker | 47% | 53% | 15 |
| quarter panel | 58% | 42% | 31 |
| door shell | 63% | 37% | 93 |
| windshield/glass | 66% | 34% | 38 |
| bumper cover | 72% | 28% | 118 |
| fender | 72% | 28% | 61 |
| hood | 81% | 19% | 37 |
| taillamp | 82% | 18% | 45 |
| roof | 89% | 11% | 9 |
| headlamp | 89% | 11% | 37 |
| moulding/trim | 93% | 7% | 104 |
| wheel | 94% | 6% | 49 |
| mirror | 94% | 6% | 17 |
| deck lid/liftgate | 95% | 5% | 20 |
| bumper impact bar | 96% | 4% | 25 |
| grille | 96% | 4% | 55 |
Framing decides what the estimate can claim
The instinct is to photograph the damage. The photograph that carries weight is of the damage and the two panels either side of it, because the argument is almost never about the dent. It is about what the dent implies for the panel next to it. This site keeps the shot list itself on a separate page rather than repeating it here.
- A close-up with no context is a photograph of a scratch on an unidentified vehicle
- Panel gaps in the wide shot are the cheapest evidence of structural movement there is
- Photograph before anything is touched: once the cover is off, the condition it arrived in is a memory
Where the photograph stops and the supplement starts
Hidden damage found at teardown is what the supplement process exists for, and no photograph and no software removes it. What a good first pass removes is the other kind of supplement, the one for something that was visible in the first frame and never got written. Those are the ones that cost credibility on the lines that are real.
How this was measured
- Panel decisions come from 912 lines in carrier estimates whose operation was R&R or repair, matched to a panel by description and counted. Panels seen fewer than 6 times are excluded.
- The published file holds a replace rate and a sample size per panel and nothing else. No line from any individual claim is in it, which is what makes publishing it safe.
- The concealed-components grouping is editorial. Whether a camera can see a part is a fact about the car rather than about the data, so that list was written by hand and the measured rates were then attached to it.
- The same table is what the estimator itself is grounded on, so what is on this page is what the product uses rather than a marketing restatement of it.
What this page cannot tell you
Published for the same reason the sample sizes are. A guide that never names its own limits is asking to be trusted rather than checked.
- Several panels sit at exactly the published minimum of 6 observations. A panel that was replaced 6 times out of 6 is consistent with a real pattern and also consistent with luck. Read those rows as a direction, not a rate.
- Nothing here measures photo estimates against teardown estimates on the same vehicle. That is the study that would actually answer the question this page is named after, and it does not exist yet.
- The corpus records what a carrier decided, not whether the decision was correct. A panel that adjusters repaired 60% of the time is a fact about adjusters, and a shop that believes those calls were wrong is arguing with the industry rather than with this table.
- There is no accuracy percentage for photo-derived estimates on this site, because none has been measured.
Before you ask
Can an estimate written from photographs be final?
Rarely, and not because of the photographs. Concealed components are the most reliably replaced parts in this corpus and none of them is visible until the cover comes off. A photo estimate that reads as final on a front-end hit is making a claim about parts nobody has seen.
How many photographs are enough?
Enough to establish the vehicle, the damaged panel, the panels either side of it and the break line itself. The count matters less than the coverage: four deliberate frames in the same order every time beat fifteen taken at random.
Should hidden damage be allowed for in the first estimate?
Not as an invented line. Note where hidden damage is likely and price it when it is found. An allowance for damage nobody has seen is the fastest way to lose the argument on the lines that are real.
Does this corpus prove photo estimating works?
No, and it is not offered as proof. It shows which decisions the corpus settles unanimously and which it splits, which tells an estimator where a photograph is sufficient evidence and where it is only supporting evidence. That is a narrower and more useful claim.
Check the evidence
Nothing on this page is worth more than what it rests on. These are the pages the figures came from.
The data
Repair or replace, measured
The full panel table these rates are drawn from.
A real repair
2017 Porsche 718 Boxster S, Progressive
A front-end hit where the whole disagreement sat in three rows of one column.
Written for
For independent appraisers
Field workflow, and what a documented decision is worth in a dispute.
Related
How to photograph collision damage
The four shots, in order. The shot list this page deliberately does not repeat.
Related
Photo estimating in the product
What happens to the photographs once they leave the phone.
Related
The iPhone app
Where the photographs are taken, and what the phone checks before it sends them.
Run it on damage you have already written
Accounts are created by hand, usually within a business day. Bring your own photographs and a repair order you have already closed, so you can see how close it lands before you decide. No card, no trial timer.