For the person who has migrated off a supplier before
Records without the relationships between them, attachments without what they were attached to, a schedule without the rules that produced it. Each is a complete export and none of them lets anybody stand the operation back up. This page describes what leaving actually requires, and commits to it while you still have leverage.
Anybody who has moved an organisation off a supplier reads the termination assistance section before the commercial terms, because they have learned where the pain actually lives. The clause is usually short, drafted by the party with no incentive to make it work, and agreed by somebody enthusiastic about starting rather than about finishing.
The export arrives and it is complete. It is also a folder of files whose relationships have been flattened away — records that referenced each other now reference nothing, attachments detached from what they belonged to, a history of who did what reduced to a column of names, and a set of rules that governed the operation preserved nowhere at all because they lived in configuration rather than in data.
And the leverage is inverted at exactly the moment it is needed. During procurement you can ask for anything and get most of it. During termination you are asking a supplier you have just told you are leaving for extra effort, on a deadline, while their team is being reassigned. Every experienced person knows the only reliable exit provision is one negotiated before signature.
Then there is what nobody exports because nobody wrote it down. Why that queue is ordered the way it is, which category means what, which exception exists because of a regulator letter three years ago, and which rule everybody works around. It is real operational knowledge, it lives partly in the supplier’s team, and it is not in any format.
The final indignity is the timeline. Notice periods assume a migration is a project with a start; in practice the receiving system is chosen late, the mapping work is discovered during rather than before, and the access is withdrawn on the contractual date regardless of where the work has reached.
A supplier can satisfy every word of a termination clause and leave an organisation unable to operate — because the clause asks for the records and the organisation needs the records, the relationships between them, the rules that governed them and the history of what was done.
So the commitment here is drawn around reconstitution rather than around export. What is available on demand, at any time and not only at termination, is the complete operating record with its relationships intact — records that reference each other still referencing each other, attachments still attached, and the history of who did what to which record preserved as history rather than flattened into a column.
The configuration goes with it, and this is the part most exports omit entirely. The rules that governed the operation — how work was categorised, how it was routed, what the thresholds were, which exceptions existed and what triggered them — are exported as a readable description rather than left behind as the reason the receiving system behaves differently. An organisation that receives records without rules has received the easy half.
It is self-service and continuous, which is the property that actually matters. The export is something your team runs whenever it wants, on a schedule if you prefer, without a request, a ticket or a notice period. That single design decision removes almost all of the leverage problem, because the artefact you need at the worst moment is one you already have from last week.
Formats are chosen so that a receiving system can consume them without a bespoke project: open, documented, with a data dictionary describing every field in ordinary language rather than in ours. A format only we can read is a lock disguised as a file.
And the honest limit, stated because your review will find it: what cannot be exported is the operational knowledge nobody wrote down. What we can do is make that smaller — the rules are in the export rather than in somebody’s head, and the reason an exception exists is recorded beside the exception. What we cannot do is claim that everything a supplier’s team knows is transferable, because no supplier’s is.
Whether an export is self-service — measured by running one in a trial without asking anybody, and timing it.
Whether relationships survive the export — measured by loading the export elsewhere and confirming records still reference each other.
Whether attachments stay attached — measured by checking a sample of attachments against the records they belonged to.
Whether the operating rules come out — measured by finding the categories, routing and thresholds in the export rather than in a screen.
Whether a receiving system can consume it — measured by a reconstitution rehearsal into any other system, performed before signature.
What access exists after notice — measured by the assistance period written in the agreement as a number of days.
no claim that operational knowledge held in people is exportable, because no supplier’s is. No claim to convert the export into another supplier’s format — the formats are open and documented and the mapping is yours or your next supplier’s. No claim that a reconstitution is effortless; it is a project, and the commitment is that it is a possible one. An independent SOC 2 Type II attestation is in progress and no report exists yet.
Where connections run under your own accounts — messaging, payments, model providers — those relationships do not move when you leave, because they were never ours. That is a significant and frequently overlooked part of an exit: the parts of an operation that live in your own supplier contracts continue uninterrupted while the rest is being reconstituted.
Your log stream already holds its copy of what happened, so the historical record does not have to survive an export at all for the part that already left the building. An organisation whose only history lives inside the supplier it is leaving is in a materially worse position, and it is a position that is entirely avoidable in advance.
And where the deployment runs inside infrastructure you already operate, exit is mostly a question of what you keep running rather than what you extract. The records are already in a database you administer, and the reconstitution question narrows to the software rather than the data.
There is no commercial argument for making leaving easy, and pretending otherwise would be the least believable sentence on the page. The actual argument is narrower and it is honest: an organisation that cannot leave never fully commits, and a supplier who has to rely on the difficulty of leaving has stopped competing on whether the work is worth paying for.
The version of that argument you can check is the self-service export. A commitment to assist at termination is a promise about future behaviour under adverse conditions; an export your team can run today without asking is a property you can verify this afternoon, and it does not depend on anybody’s intentions at the moment of departure.
The limit is stated rather than glossed: operational knowledge held in people does not export, from any supplier. What is done about it is making the surface smaller — rules in the export rather than in a screen, and the reason an exception exists recorded beside the exception — and that is a real reduction rather than a solution.
Every property here is designed rather than independently attested. A SOC 2 Type II attestation is in progress and no report exists yet; it is an attestation with a scope and a period rather than a certification. Nobody has independently examined the export claims, which is a reason to rehearse a reconstitution yourself rather than a reason to accept or reject the page.
Run the export during the trial and try to load it somewhere else. Not read it — load it. That is the difference between an export and a reconstitution, and an afternoon spent on it during evaluation replaces a clause you would otherwise be relying on at the worst possible moment.
Check specifically that records still reference each other and that attachments are still attached, because those are the two properties that most often do not survive and that a file listing cannot reveal.
Put the assistance period in the agreement as a number of days after notice, and make the export right independent of the contract state. An export that stops working the day you give notice is the one arrangement that guarantees the migration happens under pressure.
And schedule an export into your own storage from the first week rather than at the end. It costs nothing, it makes every subsequent question about exit hypothetical, and it is the only exit control that works regardless of what any supplier does.
One full export during the trial, loaded into any other system by your team, before the commercial conversation reaches terms.
Rehearsing the exit first inverts the usual order for a reason: it is the only moment when you have both the leverage to demand it and no pressure to accept whatever arrives. Everything after it is negotiated from a stronger position.
Loading it somewhere else is the whole test. A file you can open is not a reconstitution, and the properties that fail — relationships, attachments, rules — only fail visibly when something tries to use them.
Then set up the scheduled export into your own storage in the first week of the arrangement rather than at the end of it. From that point on, every exit question is hypothetical rather than urgent.
They do, and it is a promise about future behaviour by a party you have just told you are leaving, on a deadline, while their team is being reassigned. That is the weakest possible form of a commitment. The version you can actually verify is the self-service export: your team runs it today, without asking, and can run it every week thereafter into storage you control. At that point the assistance clause stops being load-bearing, because the artefact you would need at the worst moment is one you already have from last week.
That is the normal outcome and it is why this page distinguishes export from reconstitution. What usually fails is not completeness — it is that the relationships between records were flattened, the attachments were separated from what they belonged to, and the rules that governed the operation lived in configuration and were exported nowhere. All three are in this export by design. Do not take that on trust: run one during the trial and load it into something else. Reading it proves nothing; loading it proves everything.
That does not export, from us or from any supplier, and a page claiming otherwise would be lying about the part that hurts most. What is done about it is reducing the surface rather than eliminating it: the categories, routing, thresholds and exceptions are in the export as a readable description rather than living in a screen, and the reason an exception exists is recorded beside the exception rather than in somebody’s memory. That is a real reduction and it is not a solution, and you should size the residual accordingly.
Yes, and that should be written into the agreement rather than assumed, with every supplier. An export right that depends on the contract being in good standing is the arrangement that guarantees your migration happens under pressure, at the exact moment your leverage is lowest. Ask for it as an unconditional right plus a stated assistance period in days after notice — and then make it moot by scheduling exports into your own storage from the first week.
There is no commercial argument for it and pretending there is would be the least believable sentence here. The honest argument is narrower: an organisation that cannot leave never fully commits, and a supplier relying on the difficulty of leaving has stopped competing on whether the work is worth paying for. You do not have to believe any of that, because the claim is checkable this afternoon — run the export in the trial, load it elsewhere, and judge the argument by whether it worked.