How it learns

An estimator that gets graded by the carrier, and learns from the grade.

EstimateLogic Pro is compared against the carrier’s own estimate on real cars, line by line, and what it missed is what it reads the next time it writes one.

carrier estimates read line by line: 252
carrier line items: 9,434
carrier-paid part prices: 2,033
cars matched to the carrier’s own estimate: 32

The loop

1
The carrier writes its estimate
When an adjuster writes the estimate of record on a car we have already estimated, that document is the answer key.
2
It is compared line by line
Same VIN, same car. Every operation the carrier wrote is checked against what EstimateLogic Pro wrote: which lines it missed, which it priced long or short.
3
What it missed becomes what it reads
The parts prices, labor allowances and its own measured misses are rebuilt from the enlarged set. A third of the matched cars are held back, so every change is measured on cars it never learned from.
4
The next estimate is written against it
Twice a day, with no app update and no release. The estimate a shop writes this afternoon reads what the carriers wrote this morning.

What it has learned

Each change is dated, and each was measured against the carrier’s estimate on the same cars before it shipped.

2026-09-01

Paint hours

Found.: Refinish hours ran 110% over the carrier on six of six cars. The cause was a rule of thumb, not judgement: it had been told 8 to 10 hours for a door.
Changed.: The rule was deleted and replaced with the refinish time carriers actually allowed, panel by panel, measured across settled estimates.
Result.: Refinish error 110% down to 29% on the same six cars, with almost no bias left.

2026-09-01

What has to come off

Found.: Where it wrote the same operation as the carrier the hours already agreed. It was missing whole lines: the mirror, tail lamp and liner that come off to reach the damage.
Changed.: It now reads how often carriers write each removal-and-install line, and checks its own list against it.
Result.: Removal-and-install lines went from 52% to 78% of the carrier’s on six cars.

2026-10-02

Glass labor

Found.: Two cars priced no glass time where the carrier did, one of them while its own notes described removing the windshield.
Changed.: Glass operations got their own measured table, read off carrier estimates.
Result.: The windshield job that priced 0.0 glass hours wrote up to 1.2 when replayed. Not yet on every run, and the page will say so until it is.

2026-10-17

Bumper overhaul

Found.: On 19 cars from one shop the carrier wrote 119 more body hours than the estimator. About 32 of them were one line it could not write: overhauling the bumper assembly, which moves every bracket and sensor to the new cover.
Changed.: Overhaul is now an operation it can write, at the time carriers measurably allow for it, with the replaced cover carrying no labour of its own so nothing is counted twice.
Result.: On 10 cars it had never learned from, the carrier wrote a bumper overhaul on 9. The estimator now writes it on all 9, at a median 3.5 hours against the carrier’s 3.2, and its body hours rose from 59% to 68% of the carrier’s.

2026-10-17

Its own mistakes

Found.: Every table it read described what carriers do. None described what this estimator gets wrong.
Changed.: It now reads a list of its own measured misses, rebuilt from every car later matched to the carrier’s estimate, and a third of those cars are held back so the effect can be measured on cars it never learned from.
Result.: It now writes the clear coat line it used to leave off, on 4 of the 7 held-back cars where the carrier wrote one. Each new batch of carrier estimates rebuilds the list and is its next test.

Overhaul times are read from 63 settled carrier claims that wrote one.

Graded in public

The scoreboard is published car by car: 31 vehicles against the carrier’s own estimate of record, every row shown, including the worst. A tool that learns from its grade has to show the grade.

See every car on the scoreboard

Why it compounds

The answer key is the carrier’s own document on a car the estimator already wrote. That only exists where the estimator has done real work on a real claim, so every car that is later matched to its carrier estimate makes the next one better. It is not something that can be copied by reading this page.

Put it on your next car

Photograph the damage, get the estimate. Start free, no card.

Start free