The short answer
A repair order records a symptom, a trained technician's diagnosis and the fix, which is the kind of real working record AI training-data companies say they want. Nyne lists service histories among the records it takes. The value is in the cause and correction notes; the customer details and VINs have to come out before anything leaves the store.
On this page
Every service drive produces a stack of case files nobody outside the store ever reads. Each one says what the customer noticed, what the technician tested, what failed and how it was fixed. AI companies that train models to reason through real work have started paying for records like that, and a busy service department has years of them.
What a repair order actually contains
Strip away the totals and a repair order is a short story told in the same order every time:
- Concern. The customer’s words, written up by the advisor: “clunk over bumps, front right.”
- Cause. What the technician checked and found: measurements, codes, a worn sway bar link.
- Correction. What was replaced or adjusted, and the result of the road test.
- Labor and parts. Operation codes, flagged hours, part numbers.
- Vehicle. Year, make, model, engine, mileage.
Most of the value sits in the middle two lines. A model trained on thousands of real diagnoses, across makes and years, learns how experienced people narrow a problem down. That is a different thing from a repair manual, which describes how the work should be done rather than how it was.
Do buyers actually want them?
No buyer publishes a dealership program, but their published lists overlap with what a service department holds.
| Buyer | What its own page lists | Published minimum |
|---|---|---|
| Nyne | Service histories, scheduling and dispatch, CRM, support tickets | 20+ full-time on US W-2 payroll; 3 years trading |
| micro1 | Operating records, written procedures, process docs, CRM and sales history | A full-time headcount of 30 |
| Scale AI | ERP and CRM history, documents with metadata, support history | 40 staff preferred, smaller reviewed |
Taken from each buyer’s site on 7 November 2026. A buyer in this table could pay us a referral fee if a store we introduce signs with it, and that possibility has no influence on the rows.
The nearest reported deal is in a neighboring trade: a 2026 feature in The Information described labeling startups buying HVAC contractors’ records for around $150,000. A service department’s records are similar in kind: field work, written up by technicians, over years. Read that as one reported example and not a price for your ROs; and actual licenses are often far lower; headline figures travel further than typical ones.
What makes one store’s history better
Size helps, but buyers’ reviews turn on quality. Four things to check before you talk to anyone:
Narrative depth
Pull a hundred closed ROs at random. If the cause and correction lines are full sentences with test results, you have something. If most read “performed service” or “replaced part”, the dataset is thin no matter how many years you have.
One system, many years
micro1 counts a long record history in a seller’s favor. Records split across a DMS conversion can be harder to use. Find out where your pre-conversion history lives and whether you can still export it.
Breadth
Multiple makes, warranty and customer pay, diagnosis as well as maintenance. A quick-lube heavy history is less interesting than a mix of difficult work.
Clean fields
Customer details in dedicated fields can be dropped by column. Customer names typed into technician notes need a text pass.
Preparing an export without exposing anyone
- Decide the scope. Which stores, which years, service only. Leave F&I and HR out entirely.
- Drop the customer fields. Name, phone, email, address, signature, payment details.
- Drop or truncate the VIN. It links the record to a registered owner.
- Replace employee names with codes. A stable code keeps the pattern of who did what without the person.
- Read the free text. Search technician and advisor notes for names, phone numbers and addresses.
- Check the warranty lines. Warranty claims also go to the manufacturer under your sales and service agreement; read it, or leave claim detail out.
- Have someone else check it. A second person reviews a sample of the finished file.
Buyers scrub data as well; Scale AI does it on your own systems before extraction. That is the second layer of protection, not the first.
Before any of this: the contract
The records are yours, but the system holding them is not. Your DMS agreement may limit how bulk exports are made, what they cost, and whether an outside party may work inside your environment. Read it before the first conversation with a buyer; our guide to DMS contracts and data ownership lists the seven clauses that matter.
Where to go from here
If your history passes the hundred-RO test and your contract allows an export, the next question is which buyers your store qualifies for. The overview at sell dealership data sets out the published ranges and what has to stay home, and the field-by-field table on the service repair orders page is a good thing to hand your service manager.
The best technicians in the building have been writing this dataset for years without knowing it; whether it leaves the store is your decision.