We only back companies their customers could leave

Ask most software investors what makes a business defensible and switching cost will be in the first three answers. It is measured, optimised and celebrated. Data gravity, deep integrations, proprietary formats, workflows that took a year to configure — all of it described as a moat, which is a word for something that stops people leaving.

The position here is narrower than an objection to that: switching cost is real and it does work. The bet is that its value is falling and its risk is rising, and that companies built on it will be repriced before the people relying on it expect.

Why it worked for so long

Because leaving genuinely was hard, and the alternatives genuinely were worse. Migrating meant rebuilding integrations by hand, exporting into formats nobody else read, retraining everybody and accepting months of degraded operations. A customer who disliked their vendor still stayed, because the arithmetic said so.

That arithmetic depended on a few things being expensive: moving data, reimplementing logic, understanding an unfamiliar system, and finding somebody who knew both sides.

What is changing underneath it

Migration got cheaper. The work of translating between schemas, rewriting integration glue and reading unfamiliar systems is exactly the kind of work that has become dramatically less expensive in the last few years. A moat made of tedium erodes when tedium gets cheap.

Buyers learned to check. Exit terms, export formats and data portability now appear in procurement questions from organisations that would not have asked five years ago. Not everywhere, and not always well, but the direction is one way.

Regulation moved. European rules on data portability and cloud switching push in the same direction, and while the practical effect is slower than the drafting suggests, it changes what a buyer can demand and what a vendor can refuse.

Concentration became visible. Enough organisations have been surprised by a price change, an acquisition or a discontinued product that dependence now reads as a risk to be reported rather than an efficiency to be assumed.

What it means for a company being built now

A business whose retention comes from the difficulty of leaving is holding an asset that is depreciating and a risk that is appreciating. The revenue looks identical either way until the moment it does not.

A business whose customers could leave has to keep earning them, which is harder and produces a better company. It also has a more durable position when the comparison stops being feature lists and starts being what the customer owns.

What we actually check

Can the customer run it without you? Not in theory. On their own infrastructure, with their own credentials, using documentation that exists.

What does an export contain? Everything, in documented formats, including the parts that are inconvenient to hand over — configuration, history, the reasoning behind decisions.

Is the exit written down? In the contract, agreed at signature, with a defined window and a defined cost. A supplier unwilling to write it has told you where their retention comes from.

Does the licence match the marketing? We read the licence first, because it is the only part that is enforceable.

The part we could be wrong about

Buyers might keep choosing convenience. Most do, most of the time, and a thesis that requires customers to act on principle is a bad thesis.

The claim here is not that buyers become principled. It is that leaving gets cheap enough that staying becomes a real choice, and vendors relying on it being impossible lose the difference. If migration stays expensive and procurement stops asking, the thesis is wrong — and that would be visible in the market long before it was visible in a portfolio.