A Trial Period You Direct Is Employment, Whatever the Contract Says

Which test will you actually be judged against?

More than one, and they do not agree with each other. The IRS weighs behavioral control, financial control and the type of relationship, and states outright that no set number of factors decides it and no single factor stands alone ¹. Federal wage law asks a different question, whether the worker is economically dependent on you, under an analysis the Department of Labor proposed rewriting again in February 2026 ². Several states go further still.

The state tests are the ones that bite a small company, because two of them invert the burden of proof. California presumes that anyone providing labor for pay is an employee unless the hiring entity demonstrates all three of the following: that the person is free from its control and direction in fact as well as under the contract, that the work sits outside the usual course of its business, and that the person is customarily engaged in an independently established trade of the same nature ³. California also exempts a list of occupations from that test, mostly licensed professions including lawyers, accountants, architects and engineers, which are judged under an older multifactor standard instead ⁶. Massachusetts runs the same three-part structure ⁴.

One detail in the Massachusetts statute is worth reading twice, because it disposes of the whole paperwork theory of classification. Failure to withhold income taxes, or to pay unemployment contributions or workers' compensation premiums, is expressly not to be considered in making the determination ⁴. Sending a 1099 instead of a W-2 is not evidence that somebody is a contractor. It is a consequence of a classification that was already either right or wrong.

This is public legal fact rather than advice, the federal analysis is mid-rewrite as of the 2026 proposal, and the tests differ by state and by statute within a state. Before an offer goes out, put the specific arrangement in front of an employment lawyer for an hour. That hour is cheaper than a reclassification, and it is the one professional service a first-time hirer should not skip.

Why does a trial month fail the test it is meant to pass?

Because direction is the axis every one of these tests turns on, and direction is the entire appeal of a trial. You want to set the priorities, look at the work in progress, ask for changes, and watch how the person responds when you do. That is a description of supervision, and the IRS says the keys are to look at the whole relationship and the extent of the right to direct and control the worker ¹.

The second prong of a state ABC test is worse for your case, and founders rarely see it coming. It asks whether the work sits outside the usual course of your business ³. A first hire into a function you are building is by definition inside your usual course: that is why you are hiring for it. A bookkeeper or a commercial cleaner can clear that prong at a software company. The first designer at a design-heavy product company cannot, however the contract is drafted.

Third, the right to control counts even where you never use it. A contract that reserves your approval over how the work is done, or that sets hours, or that requires the person to work in your tools on your systems alongside your team, is describing an employment relationship even in a week where you left them alone. Nobody is accusing you of anything by reclassifying: the analysis looks at what the relationship was, not at what either side meant it to be.

The bill runs wider than a fine: back payroll taxes, unpaid overtime exposure, benefits, penalties, and a person who now has an employment claim against a company with no HR function and no documentation. Weigh that against the thing you actually wanted, which was three weeks of watching somebody work.

Replace the trial month with one scoped piece of paid work

Buy the evidence directly rather than through a contract that describes something else. One piece of real work, fixed fee, fixed deadline, a brief you hand over and do not supervise, judged on the artefact and on the reasoning behind it. Then make an employment offer. You get the look at real work you wanted, and none of it depends on calling an ongoing job a contract.

What makes it hold up is the absence of direction, so build that in deliberately:

• One brief, handed over complete. Questions answered in writing, no mid-flight steering, no daily check-ins.
• A fixed fee at market rate for the hours, paid on delivery, regardless of whether you go on to hire.
• Their tools, their hours, their sequence. If it has to happen inside your systems for security reasons, that is a signal worth noticing about which arrangement this really is.
• A written debrief instead of a review meeting. Ask what they would have done with another week and what they chose to leave out.

A job-shaped exercise is decent evidence, and it is worth knowing how decent. The current corrected estimate for work sample validity is .33, not the .54 that circulates, and that figure comes from studies where 53 of 54 samples tested people already doing the job ⁵. It is also not a study of unpaid multi-hour take-homes, which is a different thing wearing the same name. Treat one paid piece as one strong input among two or three, and pair it with a reference conversation. Whether the exercise or a paid trial project fits your case better comes down to how much of the job is visible in a single deliverable.

Paying somebody who is not on your payroll for a piece of work is still a contractor payment, and it gets read against the same tests ³⁴. What changes is the size of the question: days instead of a month, one defined deliverable, no supervision, no ongoing relationship, and an employment offer at the end. The version that removes the question altogether is to hire the person first and use an introductory period, which is an employment relationship from day one. Which of the two fits is a question for counsel on your own facts.

If the work genuinely cannot be scoped into one deliverable, that is your answer about classification, not a reason to write a longer contract. A body of work you have to sequence week by week is a job.

Say what the paid work is for before you send it

Tell the candidate what the piece is for in the message that carries it: this is an evaluation, it pays this much, it takes about this long, and the outcome is either an employment offer or a no. Somebody who thought they were auditioning for an ongoing contract and finds out otherwise is a candidate you lose late, after they have already turned something else down.

Say the same thing about the role itself. If you are hiring an employee, say employee in the first conversation and put the range next to it. Contract-to-hire has a specific meaning in the market, and if what you mean is a paid evaluation followed by a decision, use those words instead, because the mismatch surfaces at the offer stage where it costs both sides the most.

There is a version of this question that is genuinely a contractor question, and it is worth separating out. If the need is a defined project with an end date, if the person has other clients and their own tooling and pricing, and if you want the deliverable rather than the person, hire a contractor and stop reading. The awkwardness only appears when the true motivation is evaluation, because evaluation wants proximity and proximity is what the tests measure.

For a company with no HR function the whole hiring process should fit on about a page: the bar you set, the questions everybody gets, the paid exercise and its brief, who decides, and what you keep. That page is also your answer if the classification is ever questioned. The smallest hiring process a two-person company can defend is that page written out, and if nobody there has done the job, the standard on it should come from two practitioners rather than from a template. It takes an afternoon, and it is a better first purchase than an employer-of-record subscription.

Read the evidence >