The finance model treats them as a lower unit cost. No employer contributions, no benefits, no notice period, no severance, and the whole thing cancellable in thirty days. That model is the reason the arrangement gets chosen, and it is also the reason it is so often wrong, because the savings it counts are exactly the obligations that a regulator will later say you owed anyway.
This is not a legal briefing. It is orientation for a founder deciding how to engage the next twenty people, and every jurisdiction-specific question in it needs an answer from somebody qualified in that jurisdiction. What follows is the shape of the problem, which is remarkably consistent across countries even where the tests are not.
What the tests are actually looking at
Different legal systems ask different questions in different orders, and several run more than one test depending on whether the issue is tax, employment rights or social insurance. Underneath, the same themes recur.
Control is usually first. Who decides how the work is done, when it happens, and where. A contractor is engaged for an outcome; an employee is directed through a process. Integration is next: is this person part of the organisation, in your team structure, at your standups, with your email address and your job titles, or are they a supplier operating alongside it.
Substitution asks whether the person could send somebody competent in their place and still be paid. In a real contract for services that is often possible; in an employment relationship it is unthinkable, because you hired that person. Economic reality asks whether they carry any risk of loss, whether they can profit from working efficiently, and whether they have other clients or depend on you for effectively all their income. And the tools question asks who supplies the laptop, the software, the workspace and the systems.
Two things follow from that list. First, none of these is decisive alone. A regulator weighs the picture, which means an arrangement can survive one uncomfortable answer and fail on the accumulation of five. Second, every one of them is a fact about behaviour, not a fact about paper.
Why the heading on the contract is the weakest evidence you have
The general principle, stated in different words in different places, is that substance beats label. If it were otherwise, employment law would be optional for anyone with a word processor.
So the agreement is not worthless, but it is not what people think it is. It is useful because it records what both parties intended, sets out substitution and control in a way you can point at, and demonstrates the arrangement was designed rather than drifted into. It becomes worthless the moment daily practice contradicts it, and daily practice usually does, because the contract was written once and the working relationship is rewritten every week by a hundred small decisions nobody logs.
The practical test a founder can apply without a lawyer is uncomfortable and quick. Describe the working relationship to somebody outside the company, using only what actually happens in a normal week, and never using the word contractor. If what you have described is a job, you have a problem that the agreement will not solve.
What it costs when it lands
Misclassification is rarely a single bill. It arrives as a stack: unpaid employment taxes and social contributions across the whole engagement, interest, penalties, and retrospective entitlements the person never received, such as paid leave, notice, severance or pension. In some jurisdictions there are separate consequences on the tax side and the employment side, brought by different authorities on different timetables.
Then there is the part nobody models. These things surface during diligence. An acquirer finding forty long-running full-time contractors will not walk away, but they will price it, hold it in escrow, or require it fixed before closing, and fixing it in a hurry means converting people to employment in several countries at once on somebody else’s schedule.
The timing is the cruelty of it. The risk is created in a period when the company is small and moving fast, and it crystallises in a period when the company is doing something important and cannot afford the distraction.
The operational half, which is where the actual work is
Suppose your classification is sound. You still have not finished, because a contractor population needs its own operations, and companies routinely run it in a spreadsheet parked next to a real HR system.
Contractors need their own onboarding, and it is not the employee one with steps deleted. It is a different sequence: the contract signed, the correct tax documentation for both countries collected before the first payment, scoped access to only the systems the engagement requires, and a defined start and end. Our employee onboarding checklist is the shape of the employee equivalent, and the differences between the two are the point.
They need their own document set, with expiry dates that somebody watches: insurance certificates, registrations, permits, tax forms that change when the person’s circumstances change. They need their own money flow, which is invoices in rather than a payroll run out, with approval, rates that changed on a known date, and historical invoices that stay correct after a rate change. And they need their own offboarding, because an engagement that simply stops leaves a portal login live, an unresolved invoice, and a document set that is now retained without a reason.
The spreadsheet fails not because it is a spreadsheet but because it is a second system of record. Two systems means two answers to who is working for us this month, and the reconciliation between them is done by a person who eventually leaves. If you are choosing tooling, insist that contractors and employees live in one system with a structural separation between them, so a contractor never lands in employee headcount and never inherits employee access by accident.
Where Capstan stands
Capstan holds contractors as their own kind of record: distinct from employees, never counted in employee headcount or seat metering, with their own portal, their own document set, their own invoices and their own offboarding. That is a modest claim and it is the right one to make, because it describes bookkeeping rather than judgment. The module catalogue has the detail, and the product overview has the boundaries.
There is a classification questionnaire. It is optional, its question set is versioned content rather than opinion typed into code, and what it produces is an informational risk indication that is explicitly labelled as not legal advice. It never blocks an action, ever. We consider that a feature. A product that told you an engagement was compliant would be making a legal determination it has no standing to make, and the honest output of a questionnaire is a nudge to go and ask somebody who does.
Two things Capstan does not do, stated so nobody has to discover them later. It never moves money: contractor payments are recorded against the engagement with a reference, and the transfer happens on your own rail. And it never files anyone’s tax forms. It collects the forms and keeps them, and submitting them remains the contractor’s own business, as it should be.