How to brief and trial a freelancer

What a brief has to contain

Six things. If one is missing, you will supply it later at a worse moment.

1. The decision the work serves. Not “we need a landing page” — “people are dropping out at the pricing step and we don’t know why.” State the problem and you get somebody’s judgement attached to the work. State only the solution and you get exactly what you asked for, which is sometimes the wrong thing built well.

2. What already exists, handed over rather than described. The files, the access, the brand rules, the half-finished attempt from last year. Whatever you decide not to hand over, say so and say why, because the gap will be discovered.

3. The constraint that will actually bite. Not the whole list. The one that will: the legal review that takes a fortnight, the CMS that cannot do the thing, the stakeholder who has opinions about typefaces. Everybody has one and everybody forgets to mention it.

4. Who says yes, and how fast. Name the person. If approvals go through a committee that meets fortnightly, that is a fact about the schedule and it belongs in the brief, not in an apology in week three.

5. The real deadline, and the reason under it. Deadlines with reasons get met. Deadlines without reasons get treated as preferences, correctly.

6. What “done” looks like. The artefact, its format, where it lands. “A Figma file with the three states, delivered into this project, ready for a developer to build from” is done. “Designs” is not.

And leave out the solution where you have one but are not certain of it. Say what you were thinking and mark it as thinking. A good freelancer will tell you it is wrong. That is the most valuable thing you will get out of a trial, and a brief written as an instruction destroys it.

Choosing the trial

Pick a real slice, not a test. If you would throw the output away, it is an exam, and a good freelancer with options will decline it. Pick something small that you genuinely need — one page, one module, one week of the books, one edit — so that the worst outcome is a piece of work you paid for and can use.

Size it to a day or two. Long enough that they hit something unexpected, short enough that stopping is cheap. Most of the signal arrives in the first hours anyway, when the questions come.

Pay for it, at their stated rate, agreed in writing before it starts. Never ask for unpaid spec work. Beyond the ethics, it is a filter that selects against exactly the people you want: somebody with a full pipeline does not audition.

Give them the same brief you would give a hire. A trial run on a deliberately vague brief tests mind-reading. That is not the skill you are buying.

What to grade when it comes back

Not just the artefact. Six things, and four of them happen before the deliverable exists:

1. Did they ask a question before starting — and was it the question you were dreading? The person who finds the soft spot in your brief in the first hour is the person who will find it in month four, when it is expensive.
2. Did they hit the small deadline? A small deadline missed without a message is a large deadline missed without a message. This is the single most predictive item on the list.
3. When they hit an ambiguity, did they flag it or guess? Guessing is the expensive habit, because a guess that happens to be right teaches you nothing and the next one will not be.
4. Did what arrived match the format you asked for? Not aesthetics. Did the file open, in the thing you said, named the way you asked.
5. Can they say what they would do differently with more time? The same self-criticism test as in vetting, now with real work behind it.
6. Did they tell you something about your own problem that you did not know? Rare, and worth a great deal. It is the difference between somebody executing your brief and somebody improving it.

The debrief: twenty minutes, three questions

Do this whether or not you are continuing.

“What was the worst part of this brief?”

“What would you need from me to do this at three times the size?”

“What did I not tell you that you needed?”

The answers are a free audit of how you commission work. Somebody who answers all three candidly is somebody you can work with when something goes wrong.

When the trial is bad

Pay in full. You bought the time; the result was the information. Refusing to pay because you did not like the output is how a reputation is acquired, and freelancers talk to each other more than clients do.

Say what did not work, in one sentence, without a paragraph of softening. “The structure isn’t what we needed and I don’t think another round gets there.” People would rather have that than silence, and you may want them for something else later.

Do not extend it to be kind. Two trials that both half-worked cost more than one that failed cleanly — in money and in the weeks you spend hoping.

After a good trial, still ramp

Make the first real engagement small, with a checkpoint in the middle that either of you can stop at. Agree what a change of scope costs before you need one. Keep the scope written down.

None of this passes through this site. There is no escrow here, no milestone account and no dispute process, and money never moves through Mainstay. The trial is the protection, and the written scope is the rest of it.

What the checks on this register do and do not cover is on what a directory cannot verify. The bar every record was read against is the standard.