Why you can always leave

Every database company says your data is yours. Most of them mean it.

Then you try to take it, and you find out what yours was doing in that sentence. It was doing legal work, not practical work. Yours in the sense that nobody else claims it. Not yours in the sense that you can pick it up and carry it out.

So you start the export. One table at a time, because that’s the only shape on offer. The attachments come out as links that resolve until one day they don’t. And the relationships between your tables — the thing that made it a database instead of a pile of spreadsheets — don’t come out at all. You get the nouns and lose the grammar.

Somewhere in the third hour it lands: you don’t own this. You have visitation rights.

Nobody decided this

That’s the part worth understanding, because it’s the part that tells you it’ll happen again wherever you go next.

There was no meeting where someone said make the export bad. What happens is quieter. The export is the only feature in the entire product that exclusively helps people who are leaving. It has no champion. Nobody files a request for a better one, because the people who need it are already half out the door and have stopped filing anything. It cannot win a prioritisation meeting, because there is nothing to weigh it against that doesn’t beat it.

So it stays at the bottom of the list. Not forever by decision. Forever by default.

And underneath that, gravity

A company whose customers can’t leave doesn’t have to be as good.

It never has to decide to get worse. It just stops having to be better. Prices drift up because nothing stops them. The grid gets slower because the person who’d have walked over it can’t. Things get built because they’re easy rather than because they’re needed. None of this requires a villain — it only requires an absence. Switching costs are a cushion, and cushions make you soft.

That’s the thing we were actually frightened of. Not being bad now. The mechanism that would make us bad later, gradually, without anyone here ever noticing it happening.

So we built the exit first

Not as a kindness. As a constraint.

If you can leave any day, with everything, then every day you stay is a day we had to earn.

There’s no cushion under us. If we get slow, you go. If we get expensive, you go. If we get boring, you go. The mechanism that makes companies soft has nothing to grip.

It’s the same reason you’d tie yourself to a mast. Not because you don’t trust the sea. Because you don’t trust yourself.

The obvious objection

If we believed all of this, we’d let you run it on your own server.

That’s fair. It’s the strongest version of the argument, and we don’t do it. Self-hosting is a second product with a second support burden and a second set of ways to fail at three in the morning, and we’d rather do one thing properly than two things badly. If that’s what you want, use NocoDB — it’s open source, it’s free forever on your own machine, and we mean that without a note of sarcasm in it.

What we can do is the honest next-best. You can’t run it yourself. You can take all of it and go run something else. It isn’t the same door. It’s still a door.

What that looks like

One click. One file.

CSV, because Excel, Postgres, pandas and a text editor have all understood it for thirty years. The attachments as bytes, not links. A manifest with enough in it to rebuild the thing somewhere else.

Nothing in that zip needs us to open it. That’s the only test that matters: if you need our software to read your export, it isn’t an export. It’s a hostage photo.

It works on the free plan. It works if you’ve stopped paying us. It works on the day you’re leaving for a competitor — which is the only day it really counts, and the reason most exports are quietly bad.

We’re not doing this because we’re good. We’re doing it because we don’t want the option of getting worse.

We’d rather you could leave, and didn’t.