Guide
How a collision estimate is actually built
An estimate is four decisions repeated down a page: what is damaged, whether it is repaired or replaced, what the part costs, and what an hour of labor is worth. Three of those four are things this corpus has measured.
Last reviewed . Every figure below is read from the measured corpus at build time, so it moves when the corpus does.
Carrier line items read
6,874
from 169 parsed carrier estimates
15 named insurers produced a parsed estimate: AAA, Allstate, Bristol West, Direct General, Esurance, GEICO, Kemper/Infinity, Liberty Mutual, Mendota, National General, Progressive, Shop estimate, State Farm, USAA, Unknown.
Body labor rate band
$51 to $60
across 8 carriers, 89 documents
Weighted entirely by Progressive, which supplies 59 of those documents. Bristol West supplies 1.
Decisions that went to replacement
83%
of 912 adjuster decisions across 20 panels
The four decisions
Every line on a collision estimate is one of these, and the order matters because each one constrains the next. Get the second one wrong and the third does not exist: a repair line carries no part at all, so choosing repair where the carrier expected replacement does not make the part cheap, it removes it from the document.
- Identify the vehicle to the trim, because the part number follows the trim and not the badge
- Decide repair or replace, panel by panel
- Price the part, from what somebody actually paid rather than from a list
- Apply the labor rate that this job will actually be billed at
The labor rate is a column, not an opinion
Carriers in this corpus hold body and refinish between $51 and $60 an hour. The tightness is the finding. Progressive is the deepest sample at 59 documents and prints the same figure on every one of them, which is a contract rate rather than a negotiating position. It is not a market rate, and the clearest evidence against reading it as one is on this site: Auto-Owners Insurance allowed $125.00 an hour to the same shop in the same season. Read the document count on each row before reading the rate. One document is a single observation and this table shows it as one.
| Carrier | Body | Refinish | Documents |
|---|---|---|---|
| Progressive | $56.00 | $56.00 | 59 |
| GEICO | $56.00 | $56.00 | 8 |
| Allstate | $56.00 | $56.00 | 5 |
| State Farm | $57.00 | $57.00 | 6 |
| Liberty Mutual | $58.00 | $58.00 | 3 |
| USAA | $60.00 | $60.00 | 2 |
| Kemper / Infinity | $56.00 | $56.00 | 5 |
| Bristol West | $51.00 | $51.00 | 1 |
Hours diverge as well as rate, and that surprises people
The instinct is that a shop and a carrier disagree about the rate and agree about the work. On the 2026 Tesla Model Y published on this site, USAA paid $12,501.44. The same damage, same parts, same hours, priced at the rates measured from that carrier's own documents came to $12,272.00. Priced at the shop's posted rates it came to $13,075.00. Only the rate column moved between those two numbers, and it moved the total by more than the difference between them and the settlement. That is one job, published as one job.
Repair or replace is where the money actually is
Across 912 decisions in 20 panels, 83% went to replacement, and the split is not random. Bolt-on components are replaced almost without exception. Welded and structural panels are a genuine judgement call: the fender went to replacement in 72% of 61 decisions, the quarter panel in 58% of 31, and the rocker in 47% of 15. Those three are near enough a coin toss that how the damage is documented decides them.
What is not in this corpus at all
Nothing here measures how often an estimate was approved, how long a repair took, or what any shop billed against what it collected. Those are the numbers a shop most wants and this corpus does not contain them. It contains what carriers wrote down about parts, panels and hours, which is a narrower thing and the only thing this site claims.
How this was measured
- Documents are carrier estimates harvested from correspondence at one working shop. 239 carrier documents are on file and 169 of them are estimates a vendor parser could read line items out of, producing 6,874 lines.
- Two parsers are involved because two estimating systems are. Mitchell and CCC ONE share no format and no vocabulary, and anything compared across them has been normalised by hand.
- Labor rates are read off the rate table each carrier prints on its own estimate. The document count beside each rate is the number of separate documents that rate was seen on, so a rate measured once is visibly measured once.
- Repair-versus-replace rates come from lines whose operation was R&R or repair, matched to a panel by description, from 912 decisions. Panels seen fewer than 6 times are not published.
- Every figure is aggregate. No VIN, claim number, customer name or individual claim line appears anywhere on this site.
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.
- Bristol West rests on 1 document. That is an observation, not a rate, and it is in the table only so the reader can see the whole spread.
- Mechanical labor rates are measured badly enough that they are not published. After removing the shop's own rates from the documents, what remained read like dealer sublet rather than a carrier rate, and too few documents produced a usable figure. A third row on that table would have looked more authoritative and been worth less.
- There is no accuracy figure on this site. Not a low one, not a hedged one. The estimator has not been measured against a body of settled claims yet, and until it has, the honest number is that there is not one.
- Coverage skews hard toward European and exotic vehicles: BMW carries 293 settled part lines and Mitsubishi carries 1. Conclusions about high-volume domestic work are correspondingly weaker.
Before you ask
Is a carrier labor rate negotiable?
On the evidence here, less than shops assume. Progressive printed an identical figure on 59 separate documents, on most of them, but not all: the rest run from $49.00 to $125.00, and the high end is other shops rather than other carriers. What does move is hours and parts, and those are the columns an estimate can actually argue.
Why do these rates look so much lower than a shop's posted rate?
Because they are different things. A posted rate is what a customer paying directly is charged; the rates in the table are what a carrier reimburses on a claim. The shop behind this corpus posts $125 for body work and the volume carriers here hold to roughly half that. An estimator who prices insurance work at the posted rate produces a number that is not wrong so much as answering a different question.
Do these rates apply outside central Florida?
Unknown, and this site will not pretend otherwise. Every document in the corpus was sent to one shop in one market. Labor rates are regional, so treat the tightness of the finding as real and the level as local until somebody measures it somewhere else.
Can I see the documents behind these numbers?
No. They are real claims on real people's vehicles. The aggregate figures and the sample sizes behind them are published; the documents are not, and a vendor offering to show you somebody else's claim file should worry you.
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
Carrier labor rates, measured
The rate analysis these figures come from, with the corpus described in full.
A real repair
2026 Tesla Model Y, USAA
One job, priced twice, against what the carrier actually paid.
Written for
For independent body shops
What changes when the rate column is the thing you are arguing.
Related
Making the repair-or-replace call
The second of the four decisions, and the one that costs the most when it is wrong.
Related
The estimator itself
What the software does with the four decisions this page describes.
Related
How the damage assessment works
The model, what it is shown, and what is thrown away before you see it.
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.