| Quick facts | |
|---|---|
| The core problem | You own your records; your platform governs how you may extract and use them |
| The export path | Your own admin export of your own account — never a third-party API connector |
| Clauses that decide it | Data ownership, platform aggregate-use rights, export rights, post-termination access |
| Most common blocker | A platform term reserving AI-training rights to the vendor, not the customer |
| Most common surprise | A migration gap — history split across two systems, one of which no longer exports |
| Minimum before buyers | Three years of usable history in a system that exports at record level |
Section 1
The problem nobody mentions first
Every conversation about licensing operational data starts in the wrong place. Owners ask what their data is worth. The question that actually decides whether there is a deal is narrower and much less interesting: what does your software let you do with your own records?
You almost certainly own the data you entered. Most major field service platforms say so plainly. But ownership and permitted use are different clauses, usually in different sections, and a licensing transaction lives or dies on the second one. An operator who signs a license before counsel has read the platform terms has created a problem that is expensive to unwind.
Scale is worth keeping in view while reading the rest of this. ServiceTitan — the category leader — reported approximately 10,800 Active Customers, defined as customers with over $10,000 of annualized billings, processing $82.1 billion of gross transaction volume in the fiscal year ended January 2026, with gross retention over 95% for each of fiscal years 2026, 2025 and 2024.[1] That is the entire population of trades operators on the deepest system of record, across every trade combined. It is a small number, and it is why a well-documented file is scarce.
Section 2
What each platform export actually contains
The table below reflects what we see in practice when operators pull their own exports. It is not a vendor feature comparison — it is a read on what lands in a file a buyer can process, and where the friction usually sits.
| Platform | Typical operator | Record domains in export | Export mechanism | What to watch |
|---|---|---|---|---|
| ServiceTitan | HVAC, plumbing, electrical, roofing, multi-trade | Service calls, dispatch, estimates, memberships, invoicing, payroll joins | Account admin export / reporting module | Deepest exports in the trades. API terms restrict certain uses — read before any extraction is designed. |
| Housecall Pro | Residential service, mid-size | Jobs, estimates, customers, invoicing, scheduling | Admin CSV export | Well-structured and low-friction. Attachment-heavy records need their own export path. |
| Jobber | Smaller and newer operators across trades | Quotes, jobs, visits, invoicing, basic time tracking | Admin CSV export | Record-depth floor usually binds before the platform does. |
| FieldEdge | Longer-tenured HVAC and plumbing | Service history, equipment, agreements, QuickBooks joins | Admin export, accounting integration | Historical install bases often hold the multi-year equipment depth buyers want. |
| Aspire | Landscaping, commercial maintenance | Estimated-vs-actual job costing at crew-hour level, contracts, routes | Reporting export | The best native job-costing structure in the trades. Now inside the ServiceTitan group. |
| PestPac (WorkWave) | Pest control | Application logs, routes, recurring contracts, inspections | Admin export / reporting | Longest histories in pest. Platform-level data-use terms matter as much as export mechanics. |
| AccuLynx / JobNimbus | Roofing | Jobs, imagery, claim scope, supplements, settlements, production | Admin export; media export often separate | Bulk media export is usually the binding constraint on delivery scope. |
| Skimmer | Pool service | Reading-level water chemistry, dosing, stops, route sequence | Admin export | Confirm the export is reading-level rather than visit-summary — granularity is the asset. |
Two patterns hold across all of them. First, record-level export is the dividing line. A platform that exports reports rather than records produces something a buyer cannot train on, however rich the underlying data is. Second, attachments are where scope breaks. Inspection imagery, camera footage, signed documents and photo sets frequently live outside the main export path, and in trades like roofing and plumbing those attachments are the valuable part.
Section 3
The connector trap
There is a tempting shortcut: plug a third-party integration into your platform’s API, let it pull everything, and hand the output to a buyer. Do not do this.
The reason is structural. API and integration terms are usually a separate agreement from your subscription, they are usually stricter, and several major platforms specifically prohibit using connector access to assemble AI training data. A connector-based extraction can therefore breach terms you never read while producing a dataset whose provenance a buyer cannot verify — the worst of both outcomes.
It also fails the test that matters on the buyer’s side. A buyer subject to Article 53 of the EU AI Act must publish “draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office” and maintain a copyright-compliance policy.[2] A dataset whose extraction path is contested is a dataset they cannot document. Clean provenance is not paperwork; it is the thing being bought.
Section 4
The seven clauses that decide it
You do not need a cover-to-cover review of your platform agreement, and you should not pay for one. You need counsel to read seven things.
| Clause | What it governs | Why it matters here |
|---|---|---|
| Data ownership | Who owns customer-entered data | Most platforms affirm the customer owns their data. If yours does not say so plainly, that is the first thing to resolve. |
| Platform aggregate / derived-data rights | What the vendor may do with your records | The clause that most often blocks a deal. Many platforms reserve broad rights to use anonymized or aggregated customer data for their own products. |
| Permitted use and restrictions | What you may do with an export | Look for language limiting use to "your internal business purposes". Licensing to a third party may fall outside that. |
| API and integration terms | Separate from the main agreement | Frequently stricter than the subscription terms, and frequently prohibit building training sets through the connector. |
| Export rights and format | Whether you can get record-level data out | Summary-level export is not enough. A buyer needs records, not reports. |
| Post-termination access | What happens if you leave | Decides whether a legacy system still holds licensable history or has become unreachable. |
| Confidentiality and sublicensing | Whether you may pass data onward | A license to an AI buyer is a sublicense in substance. The terms have to permit it. |
In our experience the second row is where deals stop. A platform that has reserved broad rights to use anonymized or aggregated customer data for its own products has not taken your ownership, but it has created a question a buyer will ask: has this dataset, in some form, already been licensed into something that competes with what you are offering? That question is answerable. It is much worse when it surfaces in diligence than when you raise it yourself.
Section 5
What your platform already does with your records
This is not hypothetical. ServiceTitan’s annual report describes generating “AI-powered benchmarked industry metrics and best practices… by leveraging our most successful customers and analyzing anonymized customer data”.[1] That is a vendor building a product from the aggregate of its customers’ operational history, disclosed in a public filing.
We cite it not as a criticism — it is a common and openly disclosed model — but because it demonstrates the point concretely. Your records already have more than one party with an interest in them. Knowing which rights your vendor holds is a precondition to offering any of those rights to somebody else.
Section 6
Migration gaps and legacy systems
Most operators with real history have changed systems at least once, which creates the most common structural defect we see: a file that is nominally eight years deep but is actually two files of four and three years with a gap between them.
Three things decide whether the older history is recoverable. Whether the former platform’s post-termination terms still permit an export. Whether an export was taken at migration and still exists in a readable form. And whether the old records can be mapped onto the new schema without losing the joins — a job record that no longer connects to its estimate, its invoice and its outcome has lost most of what made it worth licensing.
Where legacy data is genuinely unreachable, we scope to the clean window and say so. A credible four-year file beats an eight-year file with an unexplained hole in it, because the hole is the first thing a buyer’s technical diligence finds.
Section 7
What to do before you talk to anybody
Four steps, in order, none of which require spending money:
- Inventory your systems, not your records. Which platforms, which years, which trades, which lines of business. Note every migration and its date.
- Locate your actual agreement — the version you accepted, not the one currently on the vendor’s website — along with any separate API or integration terms.
- Put the seven clauses in front of counsel. Narrow question, narrow bill, usable answer.
- Take a test export of a single month under your own credentials, to see what the mechanism actually produces. Do not export the full history and do not send it anywhere.
Frequently asked questions
Do I own the data in my field service software?
Can we just use an API connector to extract the dataset?
My platform already uses my data for its own AI features. Does that stop me licensing it?
We migrated platforms three years ago. Is our older history still usable?
Does it matter which platform we run?
What if our terms changed after we signed up?
Who reads the terms — you or our lawyer?
References
Sourced from primary industry reports, government data, SEC filings, and peer-reviewed valuation research. Retrieved at time of publication.
- 1.ServiceTitan, Inc. — Form 10-K for the fiscal year ended January 31, 2026 — Active Customer count, gross transaction volume, and gross retention for the leading trades system of record (2026)
- 2.European Union — Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 53 — Obligations for providers of general-purpose AI models — Copyright-compliance policy and public training-content summary; in force 2 August 2025 under Article 113(b) (2024)
- 3.U.S. Federal Trade Commission — AI (and other) Companies: Quietly Changing Your Terms of Service Could Be Unfair or Deceptive — FTC staff position that retroactively broadening data-use terms can be an unfair or deceptive practice (2024)
- 4.State of California — California Consumer Privacy Act, Civil Code § 1798.140 — definitions of "deidentified" and "sale" — Statutory standard a de-identified operational record must meet, and the breadth of "sale"