Data lock-in: You may be building the wrong kind

There are two entirely different things called data lock-in, and conflating them is how companies end up with retention numbers that look fine and a reputation that does not.

The first is friction: leaving is hard because you made it hard. No usable export, proprietary formats, history that cannot come with you.

The second is gravity: leaving is expensive because the accumulated data is genuinely more useful where it sits than it would be anywhere else.

They produce the same retention curve for a while. Only one of them survives contact with a regulator, a determined competitor, or a customer who has decided to leave anyway.

Data lock-in is switching cost created by what a customer has accumulated inside a product, not by what the product does. Features can be copied. Four years of records, decisions and context cannot, which is what makes this pattern durable.

Four things accumulate, and they are not equally hard to replace

In rising order of how hard they are to replace:

Time series is the strongest of the four, and the most commonly overlooked. A competitor can import your customer’s contacts. They cannot import three years of what normal looked like, which is one reason analytics and monitoring products tend to be stickier than their feature lists suggest.

The EU Data Act is dismantling the friction version

If your retention depends on export being painful, that dependency will not last.

The EU Data Act entered into force on 11 January 2024 and has applied since 12 September 2025. Among other things it establishes a framework for customers to switch effectively between providers of data-processing services, and voids contract terms about data access and use that have been unfairly imposed on one party.¹ It binds providers offering services in the EU market, so how much of a clock this puts on you depends on where your customers are. The direction of travel is the point rather than any single instrument: portability keeps being legislated, and each time it is, a moat becomes a compliance project.

The strategic read is straightforward. Anything you are relying on that a regulator can mandate away is borrowed time, at least in the markets where the rule applies. Gravity is not on that list, because no rule can require that a customer’s data be as useful somewhere else as it is where the history lives.

Trapped customers do not stay quiet

Friction also fails on a second count: it does not produce indifference. It produces resentment, and resentment is not neutral.

A customer who wants to leave and cannot is doing three things. They are not expanding. They are answering reference calls honestly, and in B2B those calls carry weight. And they are waiting: the moment a migration path exists, they take it, usually all at once and often loudly.

None of that is measured here, and the figures usually quoted for it do not survive checking. But it suggests a signature worth looking for in your own numbers, which is the useful form of the argument: retention holding while expansion stalls and advocacy goes negative. That combination rarely means the product is fine. It means people are staying because leaving is annoying, and you are accumulating an exit queue rather than a customer base.

Building gravity instead

Gravity comes from making accumulated data do work the customer would lose. Five moves, roughly in order of cost:

Make history do something. Data that is merely stored creates almost no gravity; data that produces a comparison, a baseline, or a trend creates a lot. The same rows become irreplaceable the moment the product uses them to say “this is unusual for you.”

Let configuration accumulate deliberately. Every view, automation and template a customer builds is switching cost they created themselves and will not enjoy rebuilding. This is also what makes a habit durable, and it is the version of accumulation most products can actually ship.

Capture what would otherwise leave with an employee. Relationship history is the clearest case: what was promised, discussed and agreed survives the person who agreed it, and a company that has been through a couple of staff changes knows exactly what that is worth. Institutional memory is gravity that the customer’s own turnover makes stronger.

Connect records to each other. A contact list is portable. A contact list where every record carries its conversation history, its owner and its outcomes is a graph, and graphs do not survive CSV export intact. Linked-note tools make the same point from the opposite direction: the files export cleanly, and what you lose is the product resolving the links between them.

Enrich what they gave you. Anything the product adds on top of the customer’s own data (derived scores, resolved duplicates, inferred structure) leaves with a much lower fidelity than it had, and the gap is visible immediately after a migration.

When nothing accumulates

The friction detector

There is one comparison on this topic worth more than all the others, and it is the one almost nobody runs: the expansion rate of your longest-tenured accounts, against your newest.

Gravity and friction both hold customers, so retention cannot tell them apart. Expansion can. A customer who stays because the product keeps getting more useful buys more of it. A customer who stays because leaving is a nightmare buys nothing further and waits. So oldest cohorts that retain well and expand poorly are the signature, and it is invisible on any dashboard that reports the two metrics separately.

Two supporting checks, both cheap. Does retention improve with tenure at constant plan size? It should, if anything real is accumulating. And how many customers use your export? Read as a churn signal, an export request is often better read as somebody establishing that they could leave, which is a precondition for committing further rather than a prelude to going.

Turning friction into gravity

1. Write down which of the two you actually have. Friction or gravity. The answer determines whether the rest of this list is urgent or optional.
2. Ship the export you have been avoiding. It is the cheapest trust you can buy, it is increasingly required anyway, and it forces the honest conversation about what really holds customers.
3. Make one historical comparison a feature. Pick data you already store and use it to tell customers something about their own past. This converts storage into gravity faster than anything else on the list.
4. Check expansion in your oldest cohorts against your newest. A gap there is the trapped-customer signature and it will not appear in your retention dashboard.
5. Find the accumulation threshold where churn drops, and make reaching it an explicit onboarding goal, not an accident.
6. Audit for the things regulation can take. Anything in your retention story that a portability rule could mandate away should be treated as revenue you are going to have to re-earn.

Accumulated data makes leaving expensive for people who stayed long enough to accumulate any. The other kind of departure belongs to the capable customer who never got that far, saw a wall of features on day one, and quietly decided this was more than they had signed up for. Deciding what to reveal and when is the difference between a product that grows with someone and one that intimidates them out of the door.

Updated August 2026