Are Your Platform Workers Actually Employees? Germany's Case-by-Case Test and the Presumption Coming in 2026

Picture a food-delivery courier or a ride-hailing driver in Berlin. They signed up through an app, their contract calls them a self-employed provider (selbstständig), they send invoices for their work, and everyone involved has treated them as a contractor. Now ask the German version of the question that Portugal and Spain have already answered: is that person actually an employee?

Unlike Portugal, Germany does not start from a presumption of employment. It has no statutory presumption for platform work, so the question is decided case by case, on how the relationship actually operates. That sounds employer-friendly, and on the surface it is. But two things make German misclassification more dangerous than a simple "it depends," and a third is about to change the starting assumption entirely. Let me walk through the test, the enforcement machinery behind it, what misclassification actually costs, and the EU directive that takes effect at the end of 2026.

No presumption, but a strict test that ignores the label

German law defines an employee in §611a of the Civil Code: someone who performs work in personal dependence, bound by instructions as to the content, method, time, duration, and place of the work, and integrated into another party's organization. The label the parties chose decides nothing. If the substance of the relationship meets that definition, the person is an employee regardless of what the contract says and regardless of the fact that they send invoices.

For platform work, courts look at exactly the features that make an app an app. Does the platform set the pay? Does it direct how the courier behaves toward the customer? Does it track the worker by GPS, rate them, and steer future work through that rating? Can the worker genuinely refuse jobs and set their own hours without penalty, or is that freedom only theoretical? The Federal Labour Court has already classified platform workers as employees where the integration was real, and in its 2020 crowdworker decision the court looked straight past the "self-employed" labels to the actual conduct: a rating-and-incentive system that in practice steered the worker's behavior was enough to make the relationship employment. Delivery-rider cases have turned on integration into the platform's operations, GPS tracking, and ratings; ride-hailing has been more mixed, depending on how much the platform actually controls.

The German model is not even uniform within delivery. The country's largest food-delivery operator long employed its riders directly as regular employees, with full social insurance, precisely to stay out of this fight; even after it began shifting much of that workforce to subcontracted fleet firms in 2025, those riders remain employees of someone rather than independent contractors. The assumption that everyone in gig work is a contractor has never held uniformly across the sector.

The status check that turns the question concrete

What makes the German test bite, in the absence of a presumption, is a specific procedure. Under §7a of Social Code Book IV, anyone, the worker, the company, or both together, can ask the Deutsche Rentenversicherung to run a status-determination procedure, the Statusfeststellungsverfahren. The procedure is free, takes around three months, and produces a decision that binds every social-insurance carrier.

This is not a rare or academic step. There were about 23,000 of these procedures in 2024, and more than 13,000 in the first half of 2025 alone. The outcomes split both ways, which is exactly the point: in the first half of 2025 the pension insurance found employment in roughly 5,100 cases and genuine self-employment in roughly 7,700. A platform cannot assume the answer will go its way. And beyond that voluntary route, customs and the pension insurance audit employers directly, and false self-employment is one of the things they look for.

The cost of misclassification: back contributions and criminal liability

This is where the German approach, no presumption but heavy enforcement, reveals itself as a deliberate design. When a worker labeled self-employed is found to have been an employee, the platform faces the full social-insurance bill for the misclassified period. Back contributions run for up to four years as standard, and up to thirty years where the misclassification was intentional. The employer owes both the employer and the employee shares, roughly 42 percent of gross across the social-insurance pillars, with very limited ability to recover the employee portion from the worker, plus a late-payment penalty of one percent per month.

Then there is the criminal exposure. Withholding employee social-security contributions is a criminal offence under §266a of the Criminal Code, punishable by a fine or up to five years' imprisonment, and the liability lands personally on managing directors and responsible executives, not only on the company. Enforcement is intense: in 2024 the customs undeclared-work unit (Finanzkontrolle Schwarzarbeit) opened close to a hundred thousand criminal proceedings across all forms of undeclared and illegal employment, false self-employment among them. So Germany never needed a presumption to make misclassification carry serious risk. It made the consequences severe enough that the "it depends" answer means real risk for anyone who guesses wrong across a whole workforce.

The presumption is coming anyway, by December 2026

Here is the third thing, and it is the one that moves Germany toward the Portuguese model. The EU Platform Work Directive, Directive 2024/2831, has to be transposed into German law by December 2, 2026. The directive introduces a rebuttable presumption of employment when indicators of control are present, and it adds rules on algorithmic management, requiring transparency about the automated systems that assign, rate, and discipline platform workers. Germany is drafting the complementary legislation and has signaled that algorithmic-transparency requirements are a priority.

Put that next to everything above and the direction is clear. Today a German platform can in principle argue, case by case, that its couriers are autonomous. From the end of 2026, once the directive is transposed, the starting assumption tilts toward employment wherever the control indicators are present, and the burden moves onto the platform to prove otherwise, which is exactly the shift Portugal made in 2023. The heavy back-contribution and criminal machinery does not go away; the presumption gets layered on top of it.

The classification call, and the payroll that follows it

The classification call itself is not one a payroll system can settle. Whether a specific courier or driver is an employee is a legal question about how that arrangement actually operates, decided on the facts and, if it comes to it, by the pension insurance or a labour court. Where a platform sets the pay, steers work through ratings, tracks couriers by GPS, and disciplines through the app, the risk of reclassification is real today and rises sharply once the directive is transposed.

What Flux handles is the half that follows the call. The moment a worker is treated as an employee, we run them as one: the five-pillar social-insurance split across health, pension, unemployment, long-term care, and accident insurance, the employer and employee shares, and wage-tax withholding, all applied correctly from the date the classification requires. Given that German back-contribution exposure reaches four years, and thirty if the misclassification was intentional, moving a worker onto correct employee payroll promptly rather than late is what limits the bill if the classification is ever challenged. Every month run correctly is a month that cannot be reclaimed as back contributions.