Ask an HR vendor whether they have a free plan and you will almost always get a yes. Ask what happens on day thirty-one and you find out which of four quite different things they meant by it. One of them is a plan. The other three are a trial, a demonstration of what you are missing, and an arrangement where the payment is made in a currency other than money.
This is a taxonomy for telling them apart, and a test you can run on any vendor in about ten minutes, including this one.
Four things wearing the same word
The trial in a plan’s clothing. Time limited, positioned in the plan table as though it were a tier, and named Free rather than Trial because the word converts better. The clock may be a date, or it may be quieter: a record limit you will hit in week three, a number of documents, a cap on stored files. The tell is a plan that appears in the marketing table but never in the contract, and copy that talks about what happens “when you are ready to upgrade” rather than what happens if you never do.
The tier built to hurt. Permanent, and deliberately shaped so that using it is worse than not having it. It holds your data but will not let you export it. It gives you one administrator, so there are no real permissions and everybody shares a login. It caps something at a number no genuine user stays under. The design goal is friction, and the giveaway is that the missing pieces cost the vendor nothing to provide. A limit that saves the vendor money is a cost boundary. A limit that saves the vendor nothing is a lever, which is the same pattern that runs the tier ladder further up the page.
The seat-capped but complete plan. The whole product, restricted on how many people you can carry. This is the honest shape, because headcount is one of the few limits in HR software that tracks a real cost: more people means more records, more documents, more storage and more support surface. It is also self-resolving. Nobody has to argue about whether you should upgrade, because the day you hire past the cap the question answers itself.
The plan where you are the product. Free because the vendor has a second buyer. Aggregated compensation and headcount data sold back to the market as benchmarks, a candidate pipeline that feeds a recruiting marketplace, or usage telemetry sold onward. This is not automatically sinister, and the better vendors disclose it properly. The difficulty is that they disclose it in the data processing agreement rather than on the pricing page, so the cost of the free plan is documented in a place buyers evaluating price do not look.
The test
Four questions, in this order. They work on any vendor and none of them require the vendor to volunteer anything.
What expires? Get a specific answer: a date, a usage total, a record count, or nothing. “Nothing” is a short sentence, so a long answer is itself the answer.
What is withheld, and what does withholding it save you? This is the diagnostic that separates the second and third categories. Ask the vendor to name the cost behind each limit. “Storage costs us per gigabyte” is an answer you can check. “That is a Professional feature” is a restatement of the question.
What happens to our data at the end? Three parts, and all three matter. Can you take a full export while still on the free plan, and in what format. What happens to the data if you simply never upgrade: retained, frozen, or deleted after a period of inactivity. And does export still work if you stop paying later, which is the version of this question that bites hardest, because it is asked at the exact moment you have the least leverage you will ever have.
Could a real company run on it? Not evaluate on it, run on it. If a genuine ten-person company could not use this as its system of record for a year, it is an evaluation environment with a friendly name. The guide to HR software for startups sets out what a system of record has to do at that size, which is the bar the free plan has to clear.
Notice that the skill of reading a pricing page adversarially does not help you much here. Almost none of these answers are on the pricing page. They are in the documentation, the processing agreement and the deletion policy, which is a fair signal in itself about which document the free plan really lives in.
Free is a distribution decision, not generosity
It is worth saying the unromantic thing, because it makes vendors easier to read. A free plan costs the vendor real money: infrastructure, support, abuse handling, and the engineering that keeps a product usable at a size that generates no revenue. Nobody does it out of kindness. It gets funded in one of three ways, by a fraction of free users converting, by your data having a second buyer, or by nothing at all, in which case the plan is a temporary marketing position and will be withdrawn or narrowed at some future board meeting.
The first is the sustainable one, and it is worth asking a vendor directly which of the three they are running. The answer, and whether they are comfortable giving it, tells you how stable the terms you are agreeing to actually are.
Where we stand
Capstan’s free plan is the third kind and we would rather be judged on the specifics than on the word.
The complete core HRIS is free up to twenty active employees. Nothing expires, there is no subscription behind it and it is never billed. The cap counts concurrently active people, so somebody who has left, or is serving their notice, does not consume a place, and it is enforced at the point an employment is activated rather than by a button in the interface. Moving to a paid plan changes exactly two things: how many people you can carry, and whether you can enable modules. It changes nothing about the software, and it does not take the twenty back either: a paid plan includes them, so the plan charge covers the employees past twenty and the twenty-first hire costs one seat rather than twenty-one. Two modules, payroll and time and attendance, come with any paid plan at no extra charge. Every other module is priced separately, sits outside that allowance and counts from the first person, because the twenty are the free core carried forward rather than a discount on everything.
Here is the limit, stated as a limit rather than buried in a footnote. A module cannot be enabled on the free plan at all. Not the cheapest one, not a single one, and not at a reduced price. If you want contractor management, performance reviews or the payroll export, you move onto a paid plan first, and that is a genuine restriction rather than a technicality. We say “the complete core”, never “the complete product”, and the difference between those two phrases is the entire module catalogue. Getting onto a paid plan is also not self-serve yet: you ask inside the app, and a person here raises the invoice.
On the third question of the test, the answers are the ones we would want from somebody else. An owner can take a full export at any time, on the free plan included, as one archive of JSON and CSV with the original document files and a documented manifest. Export deliberately sits outside the billing checks, so a workspace held for non-payment can still be exported by the people who own it. And there is no second buyer: nothing here is sold onward, nothing trains an outside model, and no software scores or ranks your people, which is written up with the architecture that enforces it on the security page.
The free plan is not a favour and we do not describe it as one. It is the part of the argument on the pricing page that we cannot walk back, renewed on every day we do not quietly gate it.