Someone good lives in Berlin. They do not want to move, you do not want to lose them, and nobody in your company has ever hired outside the home country. That is how most first international hires begin, and the structural decision that follows is usually made in an afternoon by whoever is least busy, on the basis of what a competitor appears to be doing.
There is a defensible order here, and it is not a matter of taste. Contractor first, employer of record when the relationship outgrows a contract, your own entity when the country outgrows the employer of record. Each step buys something specific and charges something specific, and the charge is mostly time and administrative attention rather than money. Skipping a step is expensive. Staying on a step too long is more expensive, and it is the mistake almost everyone makes.
This piece is about the structure. The operational half, what breaks in the second country once the hire is actually happening, is a separate list: the holiday calendar, the leave policy, the handbook, and the currency the record is kept in.
Step one, a contractor, if that is what the relationship really is
The contractor step exists because it is fast and reversible. You need a contract, a tax form appropriate to both countries, an invoicing arrangement, and a way to pay in a currency that is not yours. In administrative terms that is days of work at the start and, in a steady month, one invoice to check, one payment to make, and one document set to keep from expiring.
The condition attached to that speed is the one people skip. The step is only available if the relationship is genuinely a contract for services, which is a question of facts rather than paperwork. Our guide on contractor versus employee classification covers what those facts are, and the broader picture is in the global contractors guide. If you already know the person will work your hours, on your systems, under your direction, indefinitely, the contractor step is not a step. It is a delay with a liability attached.
Used properly, though, this stage is doing real work for you. It tells you whether the role survives contact with reality, whether the person is right, and whether the country is somewhere you want a presence at all. None of that is knowable in advance, and both later steps are much harder to unwind.
The signals that should move you off it
Five signals matter, and they are worth writing down somewhere a person actually reads.
Duration is the first. An engagement that has run past a year with no defined end is not a project. Hours are the second: full time, every week, on your calendar. Centrality is the third, and it is the one founders underrate. A contractor building a marketing site is peripheral; a contractor who is the only person who understands your billing code is core, and every classification test in the world notices the difference.
Fourth is benefits, which is a hiring signal rather than a legal one. In several countries you will simply lose candidates without local pension, insurance and statutory leave, because the alternative offers on the table have them. Fifth is headcount in country. One person is an anomaly. Three people is a site, with its own manager, its own working culture, and its own expectations about how they are employed.
What an employer of record buys, and what it takes away
An employer of record legally employs the person in their country on your behalf. They hold the employment contract, run the local payroll, make the statutory deductions and carry the filing obligations. You direct the work.
What it buys is time. You can be employing someone properly in a country in weeks, with local benefits, statutory leave and a compliant contract, without an incorporation. The monthly load is small and legible: one invoice per person, one point of contact, one set of records to keep on your own side.
What it takes away is control and, eventually, economics. The employment terms are the provider’s template, adjusted at the margins. Granting equity to someone the provider employs is awkward at best. Your own records of that person are a copy of somebody else’s system of record, which matters more than it sounds when you want to run a report across everyone. And the fee is per person per month, so the cost curve is a straight line while an entity’s is closer to a step. Two people is cheap. Twelve is a conversation with your finance lead.
Country guides are the place to check what the provider is actually absorbing on your behalf: the notice, probation and works council picture in Germany, or the contribution and pass regime in Singapore, differ enough that “we use an EOR” is not a complete answer to what you have committed to. The same holds for hiring in India, where the statutory identifiers, the notice and probation conventions and the exit process are specific enough to be worth reading against whatever the provider’s template says.
When your own entity starts to make sense
The entity step is the one to defer longest and to plan hardest. Incorporation is often the easy part. What follows is the standing load: a registered address, in some jurisdictions a resident director, a bank account, tax and social insurance registrations, a payroll registration, statutory filings on a calendar you now have to keep, and a local accountant who is not optional. That load continues in months when you hire nobody and in months when the country is quiet, and it does not go away until you formally wind the entity down, which is itself a project.
You take it on when the country is a permanent part of the company rather than a place where two people happen to live: sustained headcount, a local manager, a plan that runs for years, a need to grant equity or design your own benefits, or a provider bill that has quietly passed what maintaining an entity would cost. Any of those is a reason. None of them is urgent enough to justify opening an entity in month two.
And once you have a second entity, the HR side changes shape as well as the legal side: one directory spanning both rather than two that disagree, policy attached to an entity rather than to the company, and a transfer between them recorded as history instead of an overwrite.
The mistake, stated plainly
The classic failure is not choosing the wrong step. It is treating step one as a destination. A full-time contractor in another country, in year three, integrated into the team, on your systems, with no other clients and no right of substitution, is a permanent employment relationship that everyone has agreed to describe differently.
That arrangement does not break on a normal Tuesday. It breaks during an acquisition diligence, a funding round, a dispute with the person, or a tax authority review, which is to say it breaks at the moment you have least capacity to absorb it. The fix is unglamorous: put a review date on every long-running contractor engagement, and make somebody own the answer.
Where Capstan stands
Capstan is not an employer of record and does not sell one. It does not open entities, does not provide local counsel, and does not compute or file payroll anywhere, in any country. Those are all real parts of this problem and none of them are ours.
What it does is hold the record while you move through the steps. Contractors and employees live in one system with a structural separation between them, so a contractor never quietly becomes a line in your employee headcount, and the contractor module in the module catalogue carries the engagement, the documents and the invoices rather than a spreadsheet doing it. There is an optional classification questionnaire, and it is deliberately weak: it produces an informational risk indication, it is labelled as not being legal advice, and it never blocks an action. It is a prompt to go and ask someone qualified, not a verdict.
Two honest limits. There is no bulk importer for contractor records today, so a migration of an existing contractor population is manual work. And the free core covers up to twenty people, with everything beyond that on the published pricing page, which is the only place in this company where a number appears.