I have been in this industry for twenty-two years. I have pulled frames on everything from a 1987 Civic to a 2025 Model Y. I have watched unibody repair go from "cut the whole section out and weld in a donor" to "here is a 47-page OEM procedure for a 4-inch section."
I know the math. I know the procedures. I know the training.
And I still think we are fixing cars we should be sending to the crusher.
So here is my unpopular opinion. Try to change my mind.
The short version
Just because the repair cost is 60% of ACV and the OEM says it can be fixed does not mean it should be fixed.
Some hits change a car at the molecular level. Not just the bent parts. The metal around the bent parts. The metal that looked straight on the rack but now has internal stress you cannot measure with a tape.
We are pretending that a unibody is just a collection of sections that can be cut out and replaced like Lego bricks.
It is not.
And I have the sleepless nights to prove it.
What I have seen that changed my mind
I used to believe "if the book says yes, it is yes." That ended about ten years ago.
Case that broke me: 2015 Mazda 3. Front hit. Both rails pulled, no sectioning. Aprons straight. Strut towers perfect. Everything measured green. Customer took the car back happy. Six months later she came in with a complaint: the windshield cracked for no reason. No chip. No impact. Just a crack from the passenger side A-pillar base spreading upward.
We put it on the rack. The passenger side cowl had moved 5mm. Not enough to fail our initial post-repair measurement. But enough to twist the windshield over time. The metal had slowly relaxed into a new shape. A shape that was not straight.
We had pulled it. We had measured it. We had called it good.
And it was not good.
Second case: 2018 Camry. Rear hit. Trunk floor, rear body panel, both rear rails sectioned at the factory splice points. Beautiful repair. Welds looked like a robot did them. Sixteen months later the owner complained that the rear tires were wearing on the inside edge. Alignment was in spec. We replaced tires. They wore again. We finally put a tram gauge on the rear rail ends and found the right rail was 6mm higher at the very end than the left. The section had held. The metal forward of the section had slowly sagged.
Third case: 2020 RAV4. Side hit. B-pillar sectioned at the beltline. Rocker and roof straight. Perfect candidate on paper. At two years, the rear door on the opposite side (unhit side) started popping when you went over a driveway apron. The whole body had developed a slow twist because the repaired side was stiffer than the factory side. The car was now asymmetrical in a way no alignment could fix.
In all three cases, the math said repair. The OEM said repair. The final inspection passed.
And all three cars were permanently compromised in ways we do not have a test for.
What we do not talk about
We talk about yield strength. We talk about weld penetration. We talk about sectioning zones and cavity wax.
We do not talk about residual stress. We do not talk about metal fatigue that started the day of the crash. We do not talk about how a pulled rail that measures straight is not the same as a rail that was never bent.
Because we cannot measure those things. Not in a production shop. Not on an insurance timeline.
So we pretend they do not matter.
I am done pretending.
My personal line now
Here is where I say "total it" even when the math disagrees:
One: Any hit that requires pulling a front rail more than 6 inches AND the apron has any deformation at all. Not a tear. Just deformation. Because that apron will never be as rigid as it was. And the strut tower sits on it.
Two: Any rear hit that requires sectioning both rear rails on a unibody car. Not a truck with a separate frame. A unibody. Because the floor pan is now a hinge between two repaired rails and the car will eventually develop a diamond twist. Not if. When.
Three: Any side hit that requires a B-pillar section AND the rocker or roof had to be pulled more than 3mm. Because the pillar is the only thing holding the roof and floor together in a rollover. And I do not trust any weld to do what a continuous piece of high-strength steel did.
Four: Any hit that deformed the strut tower even 1mm. Not cracked. Not torn. Just moved. Because that tower carries the suspension load every single mile. If it moved once, it will move again.
I have walked away from jobs that hit these lines. In every case, the car went to another shop. I do not know what happened next. But I know I sleep better.
What I want to hear from you
I am posting this because I want to be wrong. If I am wrong, I can fix more cars and send fewer to the scrap pile.
So try to change my mind. Answer these:
1. Have you kept a car for five years or more after a repair that crossed one of my red lines? How did it hold up? Be honest about the problems.
2. As an industry, are we measuring the right things? If a car measures green on the rack but the windshield cracks six months later, did we actually measure wrong or is the measurement itself insufficient?
3. Is there any non-destructive testing method that can detect residual stress in a repaired rail? Something we could actually use in a real shop without adding three hours of labor?
4. If you are an insurance adjuster reading this: what would it take for you to total a car that is under your repair threshold? A customer demanding it? A shop refusing? Or nothing at all?
What I am not asking
I am not asking about the legality of totaling a repairable car. I know insurance contracts and state laws.
I am not asking about whether OEM procedures are sufficient. They are the best we have.
I am asking about the gap between what we can measure and what we should accept.
And whether that gap is big enough to change how we make the call.
A request for long-term follow-ups
If you are an owner who has kept a car for three years or more after a major unibody repair, post your story. Especially if:
The car measures straight but eats tires
The car has no visible issues but the owner says "it just feels different"
The car developed a crack or pop somewhere far from the repair
The car was hit again and you saw how the repair held (or did not)
I want data. Not theory. Not "the book says." Data from the real world.
No notes yet — write the first.