The question comes up at the worst time. A price increase lands, or the product gets acquired, or the one person who understood the system leaves, and somebody asks whether the eleven years of records inside it can be taken somewhere else. The honest answer is almost always yes, with conditions. The conditions are what matter, and they are usually written down somewhere nobody has opened since signing.
What follows compares the four routes out, on the terms that count in year eight rather than week one: what you actually receive, what it costs in ordinary staff hours, and what breaks later.
The four routes, and what each one hands you
Every business system offers at least one of these. Most offer two. Very few offer all four, and the gap between the best route a vendor supports and the one a user can click unaided is where projects stall.
| Route | What you get | Typical effort | Where it fails later |
|---|---|---|---|
| Self-service CSV or Excel export | One flat file per screen or list | Minutes per file, hours across a system | Attachments, notes, history and links between records are dropped |
| API pull (a programmatic feed the vendor publishes) | Structured records including relationships, if you build the pull | Days of developer time, then near-zero to repeat | Rate limits, and fields the API never exposed |
| Vendor-run extract | A full dataset, often a database dump or a bundle of files | A request, a fee, and a wait measured in weeks | Format is the vendor's, not yours; one shot, often no revisions |
| Document archive (PDF or print-to-file) | Readable copies of what a person saw on screen | High, unless the system batches it | Not searchable in bulk, not importable anywhere |
The distinction worth holding onto is between portable data and readable data. Portable data can be loaded into another system and worked with. Readable data can be opened and looked at by a human. A PDF archive of ten thousand invoices is readable. It is not portable. Firms discover this after they have congratulated themselves on having a copy of everything.
The paperwork that decides the answer before you ask
Four documents govern this, and they are rarely in the same folder.
The master agreement or terms of service sets out whether export is a right or a courtesy. Look for language about data ownership and a termination assistance clause. The useful version names a window, often thirty to ninety days after the account closes, during which you can still pull data. The weak version says the vendor will provide data "upon request" and stops there.
The order form is where fees hide. Some vendors charge nothing for self-service export and several thousand dollars for an engineer-run extract. That number is negotiable at signing and almost never negotiable at exit.
The data processing addendum, if you have one, usually covers deletion and return of personal information. It is written for privacy law compliance, but it often contains the only hard deadline anywhere in the stack.
The API documentation is paperwork too, though nobody files it that way. It tells you which objects and fields are exposed. If a field your business depends on is not in the docs, assume it cannot be pulled programmatically and plan another route for it.
The Federal Trade Commission oversees unfair and deceptive practices in commercial dealings, which is the broad frame these contract terms sit inside. The practical protection, though, is the clause you negotiated, not the regulator you might eventually write to.
The week-to-week version, which is the one that works
Treating export as a project you run once, under pressure, at the end of a relationship, is how firms end up with a folder of CSVs that nobody can reassemble. Treating it as a small recurring chore changes the economics entirely.
A workable rhythm looks like this. Once a month, somebody spends forty minutes pulling the three or four exports that matter: customers, transactions, and whatever the operational core of the system is. The files go to a dated folder in storage you control, not the vendor's. Once a quarter, somebody opens the oldest file and confirms it still makes sense. Once a year, somebody tries loading one of those files somewhere else, even just a spreadsheet with the columns mapped, to prove the data is portable rather than merely present.
That annual test is the part people skip, and it is the part that earns its keep. A monthly export that has been silently writing empty files since a product update in March is worse than no export, because it bought you confidence you did not have.
What ages badly, and what doesn't
Flat files age well. A CSV written in 2016 opens today. A proprietary backup format from a product that was sunset in 2019 may need the original software to read, and the original software is gone. If a vendor offers you a choice of export format, take the dullest one available: CSV, JSON, plain PDF/A for documents.
Attachments age badly for a different reason. Most systems export a spreadsheet of records with a column holding a link to each attached file, and that link points back at the vendor's servers. Thirty days after the account closes, every one of those links is a dead end. Ask specifically whether attachments come down as actual files, and whether the file names tie back to a record ID. If they do not, you have an archive of orphaned documents.
Relationships age badly as well. The line connecting an invoice to a customer to a project exists inside the system as an internal ID. Flat exports often drop it. Keeping the ID columns, even the ugly ones nobody reads, is what lets a new system stitch the pieces back together.
Choosing a route for your situation
If you are staying put and want insurance, run self-service exports monthly and test them yearly. The cost is one person's afternoon per quarter and it covers most of the risk.
If you are actively planning a move, build or buy the API pull. It is the only route that gives you repeatable, complete extracts, which matters because you will want three or four of them: one to test the import, one to fix what the test broke, and a final one on cutover weekend.
If the system is old, closed, or the vendor is unresponsive, pay for the vendor-run extract and ask in writing what format it arrives in before you commit. Getting that answer in email, with a named person attached, is worth more than the clause in the contract.
The firms that move cleanly are not the ones with the best software. They are the ones who have a dated folder going back a few years and know, because they checked in June, that the files inside it open.