The Clauses That Decide What It Costs You to Leave

The renewal quote arrived with a nineteen per cent increase. The finance director asked the obvious question — what would it cost to move to something else — and the answer took three weeks to assemble and came back as a number nobody wanted to say out loud.

Not the licence cost of the new system. The cost of leaving the old one: a data extract the vendor would produce for a fee, in a format that would need substantial work to be usable; eleven years of scanned documents that live inside the application and have no export; four integrations built by a partner who no longer exists; and a clause requiring twelve months' notice, meaning the earliest possible exit was already fourteen months away and another full year of fees.

So they renewed. And the board minute recorded that after careful evaluation the incumbent remained the best fit, which was true in the sense that it was now the only affordable option, and false in every other sense.

This is not a vendor behaving badly. Every term in that paragraph was in a contract somebody signed, at a moment when the exit was the last thing on anyone's mind. Lock-in is not something that happens to you; it is something you agree to, in writing, on a day when you are happy — and it is exactly the kind of thing an honest look at what you already own exists to surface, because it is invisible right up until the renewal letter arrives.

Lock-in is four separate things

Treating it as one concept is why it never gets addressed. It has four distinct sources, and they need different remedies.

Data lock-in. Can you get your records out, complete, in a form another system can read, without paying for a project? This is the one everybody assumes is fine and it is usually the worst.

Process lock-in. The business has reshaped itself around how the software works. Twelve years of habits, forms, reports and expectations. This is real and it is not the vendor's fault, and it is the reason a technically easy migration is still expensive.

Skills and knowledge lock-in. The three people who understand the configuration. The consultancy that built it. When the knowledge lives outside your company, changing vendor means starting again in a way that changing software does not.

Commercial lock-in. Notice periods, minimum terms, auto-renewal, uplift caps, licence counts that ratchet up and never down. These are pure contract terms, they cost nothing to negotiate at signature, they turn an eighteen-month decision into a thirty-month one, and they are the mechanism behind a large share of what shows up later as a five-year cost of ownership nobody forecast.

Only the fourth is negotiable in a meeting. The first is negotiable at signature and almost never after. The second and third are yours to manage and nobody else's.

The clauses that actually matter

Most contract review effort goes into liability caps and service levels. Those matter, and they are not what determines your freedom. These are.

Data extraction, defined in specifics. Not "the customer may request an export of its data". You want: which data — including attachments, documents and history, not just transactional tables; in what format; within how many days; at what cost, stated as a number or as zero; and how many times you may request it. The single most useful term in the whole contract is the right to a full extract at any time during the term, at no cost, at least annually — because that turns a hostile exit into a routine one, and because a vendor who will not agree to it is telling you something.

Escrow that includes the data, not only the source. Source-code escrow is a familiar term that mostly protects against the vendor disappearing. It rarely helps, because source code without a running environment and a current database is an artefact, not a system. Where escrow is genuinely needed, it should cover a periodic database snapshot in a documented schema.

Notice period and auto-renewal. Twelve-month notice with automatic renewal means that missing a date by a week costs a year. Negotiate to ninety days if you can and put the notice date in a calendar owned by a person, alongside the other dates that recur whether anybody is watching or not.

A cap on increases. An uplift capped at a stated percentage or an index. Uncapped renewal pricing is the vendor's option to reprice you once you cannot leave, and it costs nothing to close at signature.

Licence flexibility downward. Almost every agreement lets you add users. Fewer let you remove them. If your headcount can fall, and in this region it can, you want the right to reduce at renewal without penalty.

Assignment and change of control. Your vendor may be acquired. If it is, the pricing philosophy, the support model and the product roadmap can all change within a year, and you want at minimum a right to terminate without penalty on a change of control.

Documentation of configuration. A term requiring the implementer to deliver, and keep current, a written record of what was configured and why. This addresses knowledge lock-in directly and it is almost never asked for.

Interface access on standard terms. The right to use the product's own interfaces without an additional fee or a separate agreement. Where this is charged for or gated, every integration becomes a negotiation, and the practical effect is an integration you never build.

What to test before signature, not after

Contract language is necessary and it is not sufficient, because "an export in a standard format" is a phrase with a very wide range.

Ask for a sample export during the evaluation, from a real customer database with the customer's permission, or from the demonstration environment. Look at what comes out. You are looking for four things: are the attachments there; are the historical documents there; is there a schema description; and can you tell which field means what without asking. Most exports fail on the first two.

Ask what an exit actually looks like, of a reference customer who has done one. Vendors will give you references who stayed. The useful reference is one who left, and asking for that is a legitimate question that will tell you a great deal about the answer's temperature.

And ask the question in the demonstration, in front of the sales team: if we decide in three years to move, what do we get and what does it cost. An unhesitating specific answer is a good sign. A reassurance that nobody ever leaves is not an answer, and it belongs on the list of things a demonstration should be made to address.

If you are already inside it

Most readers are not at signature. They are three or seven years in, with none of the above, and the useful question is not what should have been done.

Establish what an exit would actually cost, before you need to know. Request an export now, under whatever term you have, and see what arrives. Do this while the relationship is good, because the answer takes weeks and you will not have weeks when you need it. The finding is frequently that the export is better than feared or catastrophically worse, and either changes your negotiating position immediately.

Get the documents out separately. Attachments are usually the worst part of any extract and they are also the part you can address independently, by extracting them on a schedule to storage you control. A company whose eleven years of scanned approvals, contracts and delivery notes are already in its own hands has removed the single largest exit obstacle without a migration.

Write down the configuration. Not a manual — a record of the decisions: why the chart of accounts is shaped that way, what the four custom fields are for, which reports finance actually relies on. This is a week of work by someone who was there, it decays if left, and it is worth more at exit than any contract clause.

Negotiate the terms at renewal, when you have the only leverage you will ever have. Renewal is the one moment the vendor wants something from you. Spending that leverage entirely on discount is the standard mistake; a percentage point of discount is worth far less over five years than an uncapped uplift closed, a notice period halved and a free annual export written in.

Reduce the process dependency where it is cheap to. Some of your lock-in is genuine business fit and some is habit encoded in a customisation nobody remembers requesting. The second kind is worth identifying, because it is the customisation that also makes every upgrade expensive.

The honest counterweight

Some lock-in is worth accepting, and pretending otherwise produces a different bad decision: an estate assembled from replaceable components that fit together badly.

Depth costs freedom. A system configured tightly around how your business actually runs will be harder to leave, and it will also be better while you are in it. The question is never how to have no lock-in; it is whether the lock-in you have was chosen and is proportionate to the value received.

What is never worth accepting is the accidental kind — the notice period nobody read, the export that was never tested, the documents with no route out, the configuration in one contractor's head. None of those bought you anything. They are simply the price of not having asked, and they are all fixable while things are calm.

The short version

The cost of leaving is set at signature and paid at renewal. Four separate forces create it — data, process, knowledge and contract — and only the contract half can be fixed in a meeting, which is why it should be.

Negotiate a free full extract available at least annually, including attachments and history; a capped uplift; a short notice period; the ability to reduce licences; termination on change of control; and a written record of the configuration. Then test the export before you sign, because the clause and the file are different things.

If you are already years in, do the export now while the relationship is good, pull your documents into storage you own, write down why the system is configured the way it is, and spend your renewal leverage on terms rather than entirely on discount. Accept the lock-in that came with depth you actually use, and remove the kind that came from nobody asking — because a renewal decision made under the constraint that leaving is unaffordable is not a decision, and it will be recorded in the minutes as though it were.

Where the estate is old enough that nobody can answer what leaving would cost, establishing that is a normal part of an independent look at the systems — and it is worth knowing before the renewal letter arrives rather than after.