Employer of record or entity: choose before someone chooses for you

Almost nobody plans how to employ people abroad. They solve one urgent hire, and three years later that improvised answer has become the company’s international employment structure. Fourteen contractors across six countries, none of it designed, all of it load-bearing.

For most of the last decade that worked, because nobody checked. That assumption has expired. The Netherlands resumed enforcing its rules on false self-employment in 2025 and restored penalties in 2026. In June 2026 the Dutch senate passed a law creating a presumption of employment for anyone paid below €38 an hour, with the burden of proof falling on the company. The direction of travel across the EU runs one way.

So the structure question deserves a deliberate answer. Three things decide it, and most companies never write any of them down: how long the person will be there, what the work actually looks like, and which passport they hold.

The models, minus the marketing

An independent contractor runs their own business and sells you a service. They invoice you, handle their own tax, and carry their own commercial risk. You are a client, not an employer. This remains the right structure for a great deal of work. The enforcement wave is not an argument against contractors. It is an argument against pretending.

A contractor of record sits one step further along. A provider contracts the freelancer, absorbs the classification assessment, and takes on the misclassification liability. You keep the contractor model and hand over the risk that comes with it.

An employer of record already holds a legal entity in the market and employs the person on your behalf. They run payroll, withhold tax, provide statutory benefits, and hold the contract. You direct the work and pay a fee.

Your own entity means incorporating locally and employing people directly. You become the employer, with payroll registration, statutory filings, local accounting, and liability that sits with you rather than a vendor.

These are less like competitors and more like stages. Most companies pass through several in the same market inside five years.

Question one: how long, honestly

Duration is the cleanest test, because it maps almost directly onto the models.

Work with an end date, a defined scope, and a deliverable is contractor work. A six-month project to rebuild an integration is a contract. A person who joins standup every morning indefinitely is not, whatever the paperwork says.

Companies trip on the middle case: the contractor who started as a stopgap and is still there two years later. That is not a contractor relationship that ran long. It is an employment relationship nobody converted, and it is exactly the profile enforcement targets. If you cannot name the date the engagement ends, it is not a project.

Question two: what the work actually looks like

Every European misclassification test asks variations of the same questions, and none of them concern the contract.

Does the person decide when and how they work, or do you? Do they have other clients? Do they use their own equipment? Can they send a substitute? Do they carry any risk if the work goes badly? Do they sit inside your teams, tools, and management structure?

Try this internally. Describe the person’s working week to someone outside the company without using the words contractor or employee, then ask which one they think it is. That answer is the one an auditor will reach, and the title on the agreement will not change it.

If the honest description is that they work here, the contractor option has closed. The choice narrows to a contractor of record, an employer of record, or your own entity. Continuing as before is no longer a structural choice. It is a decision to hold risk.

Question three: whose passport

This question removes the choice entirely more often than any other, and companies usually discover it last.

Sponsoring a work visa generally requires a registered local entity in good standing. In the Netherlands, only employers recognised as sponsors by the immigration service can file for the Highly Skilled Migrant permit, the EU Blue Card, or an intra-company transfer. Recognition requires a Dutch entity. No entity means no sponsorship, which means no permit, however well the interview went.

An employer of record can act as legal employer for immigration purposes. That is the only route to sponsoring a non-EU hire without incorporating first. If your shortlist includes candidates who need visa support and you hold no local entity, the decision has effectively been made for you.

What each one costs

Cost comparisons here usually cheat. They set a contractor day rate against a fully loaded employment cost, then declare the contractor cheaper. An honest comparison uses the same denominators.

Deel publishes list pricing that makes the ladder easy to see: $49 per contractor per month for contractor management, $325 per contractor per month for contractor of record, and $599 per employee per month for employer of record across 130 or more countries. Contracts run month to month.

Read those three numbers as a risk curve rather than a price list. The gap between $49 and $325 buys the transfer of classification liability. The gap between $325 and $599 buys full legal employment, benefits, and the sponsorship capability that comes with it.

What the real employment costs look like

Underneath any of these sit the real employment costs, which do not change with the model. Dutch employer on-costs reach roughly 34.25% of salary on indefinite contracts and 38.25% on fixed-term ones, plus a statutory 8% holiday allowance.

Entity setup costs less than most people assume and more to run than they budget. A Dutch BV needs minimum share capital of €0.01, a one-off KVK registration fee of €82.25, and a civil-law notary, with a realistic timeline of one to four weeks. What follows is the real expense: corporate income tax at 19% up to €200,000 of profit and 25.8% above, VAT returns, annual accounts, seven years of records, and a payroll function that has to exist whether you employ three people or thirty.

The crossover point is not a headcount figure. It is an attention figure. An entity makes sense once someone in your company can own local compliance as part of their actual job. Below that, you are buying a vendor’s attention because you have none spare, and the monthly fee is cheap for that.

The sequence that works

Test with contractors where the work is genuinely project-shaped and time-boxed, and stop when it is not. Move to a contractor of record when the relationship looks durable but the work stays project-based, or when you want the classification risk somewhere else.

Switch to an employer of record the moment someone becomes a permanent team member, and treat that conversion as a scheduled event rather than something to revisit later. Incorporate once the market is proven, headcount justifies local infrastructure, or you need sponsorship capability and control over benefits.

Plan the handover before you start it. Moving employees from an employer of record to your own entity means new contracts, continuity of service, and benefits that do not quietly reset.

Skipping stages fails in both directions. Incorporating before you know the market works leaves a dormant entity that still files returns. Staying on contractors past the point of honesty leaves an exposure that compounds until an audit or a claim surfaces all of it at once.

The reframe

The useful shift is to stop treating this as a procurement decision and start treating it as a risk decision.

Every model carries a cost and a risk profile, and the two move in opposite directions. Contractors cost least and carry the most reclassification exposure. An entity costs most to run and gives you the most control. The middle options price the transfer of a liability you would otherwise hold yourself.

Priced that way, the question stops being which option costs least. It becomes which risk you would rather own, and whether you are holding one by accident right now.