A free plan is not a smaller version of the paid plan. It is the paid plan with a wall in it, and where that wall sits decides whether upgrading is a fair trade or a toll. The useful move is to work out which wall you are heading for before you have months of work inside the tool, because that is the only point where leaving is still cheap.
The verdict, in one line
Pay when the wall you hit is retention. Leave when it's seats and your work is portable, because seat pricing scales with the thing you are trying to grow. If export fidelity is the wall, decide now — that one only gets more expensive to escape.
The four walls
When a free tier starts to pinch, the pressure usually traces back to one of these four. Whatever it limits is where the upgrade pressure will come from.
Seats — one licensed user account per person
How you notice it. A new colleague needs access and there's no room
Can you route around it without paying?. Sometimes: shared exports, a read-only view, a second workspace
Usual call. Leave, or accept a hard cap on who's inside
History / retention — how far back you can see old versions and messages
How you notice it. You go looking for last quarter's version and it isn't there
Can you route around it without paying?. Rarely — the record either exists or it doesn't
Usual call. Pay
Integrations — prebuilt connections to your other software
How you notice it. You start copying data between two tools by hand
Can you route around it without paying?. Often: manual export, a scheduled script, one connector doing double duty
Usual call. Try the workaround first
Export fidelity — how much of your work survives leaving
How you notice it. You export, open the file, and half the structure is gone
Can you route around it without paying?. No. This is the wall that stops you leaving at all
Usual call. Pay, or get out now
Seats
Seat limits are the easiest wall to see coming, and the one whose cost grows fastest if you pay past it. The cost grows with headcount, so the bill rises exactly as the team gets more valuable to you — and the pressure never lets up, because every hire renews it.
Before you upgrade for seats, check what a non-seat looks like in that product. Some tools have a free viewer or guest role, or let you publish a read-only link, and a colleague who only ever reads does not need a full account. If there is no such role, the tool has told you something: it prices per body, and it will keep pricing per body.
If seats are your wall and the work is portable, this is the moment to move. Portability is the condition — check the export before you decide, not after.
History and retention
Retention limits are the quietest wall. Nothing breaks the day you cross it. You only find out later, when you go looking for the version of a document that existed before someone rewrote it, or the thread where a decision was actually made, and the tool shows you nothing.
Two things make this worth paying for. First, you cannot reconstruct it — a retention limit is a hole in the record, and no workaround fills it retroactively. Second, the value of old history is invisible until the moment you need it, which means you will consistently underrate it while deciding.
One thing to check carefully: whether hidden history is deleted or merely paywalled. If the data is still there and only the view is locked, upgrading buys back something you already made, and downgrading later will lock it again. Ask what an export contains on the free plan, before you assume the archive is yours.
Integrations
An integration is a prebuilt connection between this tool and another piece of software. Free plans often cap how many you can run at once, which sounds severe and frequently isn't.
Work out what the integration actually does. If it moves data on a schedule you could run yourself, or if it saves a copy-paste that happens rarely, the cap is an annoyance rather than a wall. If it is doing something continuous — keeping two systems in agreement, or feeding a process other people depend on — then the manual version is a job someone has to remember to do, and that job will get skipped.
The reason this wall is different from the others: it limits convenience, not the record. Try the workaround for a while. If it holds, the cap was never the real constraint.
Export fidelity
Export fidelity is how much of your work survives the trip out. Raw text usually escapes. What often doesn't: structure, links between items, comments, attachments, formatting, version history, and anything the tool generated rather than stored.
This is the wall that decides whether the other three are negotiable. A tool you can leave cleanly has a natural ceiling on what it can charge you: past the cost of switching, leaving becomes the cheaper option. That ceiling only holds if nothing else pins you in place — a contract term, a team that would need retraining, downtime you cannot afford. A tool whose export produces a flat pile of text has you, and the price of every future plan change is set by that fact rather than by the features.
So test it early, while you have little inside. Run the export. Open the file — not the file listing, the file itself. Ask whether you could rebuild your work from it in a different product, and how much rebuilding that would be. If the answer is bad, you have two honest options: pay and treat the subscription as the cost of keeping access to your own work, or move now, when moving is small.
Predicting your own wall
You do not need to read the whole pricing page. You need one pass:
- Seats: who will need access a year from now who doesn't have it today?
- Retention: how far back would you have to look before the record stopped being useful — a week, a quarter, several years?
- Integrations: which connection, if it broke, would create work someone has to remember to do?
- Export: what does the export file actually look like? Open one.
Whichever question you answer fastest and most nervously is your wall. Start there. Check export fidelity regardless, because it decides how much leverage you have on the other three.
Who can ignore this entirely
If the tool holds work that is disposable by design — scratch notes, a one-off project, anything you'd be fine losing — free plans are simply free, and none of this applies. Same if you're evaluating three products at once and haven't committed to any: hitting a wall during a trial is useful information, not a problem to solve.
This framework matters when the work accumulates. If what's inside the tool gets more valuable the longer it sits there, the wall is worth predicting.
The one thing to do next
Open the tool you already rely on, run an export, and look at the file. Everything else — whether to upgrade, whether to move, how much leverage the vendor has over you — follows from what you see in it.