Writing a brief that gets applicants worth scoring
A brief here carries three things: the work, a small diagnostic that is a real slice of it, and a rubric saying what good looks like. The rubric is the part clients skip, and it is the part that decides whether the audition is worth anything. Here is what to write, and the mistakes that get a project removed.
Last reviewed 18 October 2026 · 6 minute read
On most platforms a brief is a wish. Here it is also an instrument, because whoever applies has to do a small piece of the work against it. A vague brief does not merely attract vague applicants — it makes the scoring meaningless, and the scoring is the product.
The short version
Write the work, a diagnostic small enough to be done in one sitting and real enough to be worth reading, and a rubric listing what you will check. Applicants do the diagnostic with their names hidden. You score every attempt against the same list, and only then do you see who made which.
The three parts
- 1The workWhat has to exist at the end, and what it is for. Written as milestones with a price each, because money here is held per milestone rather than per project.
- 2The diagnosticA small, bounded, time-boxed piece of the real work. One of five kinds: a code snippet, a system design, a piece of copy and positioning, a design critique, or a custom prompt you write.
- 3The rubricThe specific things you will check, written before anybody answers. If you cannot say what good looks like, the audition cannot be scored fairly, and it is better to know that before you post.
What makes a rubric scoreable
- Named criteria rather than adjectives. Whether an empty state is handled can be checked; whether code is clean cannot.
- The same list for everyone, decided in advance. A criterion invented after reading a submission scores that submission, not the work.
- Weights that match the job. A design-led project should not be scored mostly on documentation.
- Fewer than a dozen. Past that, nobody applies the list consistently, including you.
The line the diagnostic cannot cross
The diagnostic may not be the deliverable
Asking for the finished work as the audition is a breach of the terms: the project is removed from the directory and reviewed by hand. The task has to be a slice — small, time-boxed, and not usable as the finished thing. It is unpaid work, and the only reason it is defensible is that it is kept small.
Mistakes that cost you applicants
- No rubric. Applicants cannot tell what is being rewarded, so the field is either empty or random.
- A diagnostic that takes an afternoon. It is unpaid, and a long task selects for people with free time rather than people who are good.
- A budget with no milestones. Money is released per milestone here, so the split is part of the brief rather than a conversation afterwards.
- Skills listed instead of work described. A list of technologies does not tell anybody what to build.
- Scope hidden in conversation. Anything that matters to acceptance belongs in the brief, because whatever is not written down is what the argument is about later.
What happens after you post
- Applicants apply — one freelancer, or a squad of two to five.
- They do the diagnostic inside your window and submit it with no name attached.
- You score every attempt against your rubric, blind, and only then are the names revealed.
- You choose. Picking a lower-scored attempt is allowed if you can say why, and the scores stay recorded either way.
Two tools that do part of this
The project form offers a writer that drafts the brief from a sentence or two about what you need, and a coach that reads your draft and names what is missing, which is usually the rubric. Both produce drafts: neither posts anything, and any figure either one suggests is yours to check rather than to trust.
Questions people ask
Do I have to write a rubric?
You should, and the page will tell you when one is missing. Without it, applicants cannot tell what is being rewarded and you cannot explain the choice you make.
How long should the diagnostic take?
About as long as it takes to see how somebody thinks: one sitting, not a day. A long unpaid task selects for free time rather than for skill.
Can I ask for the finished work as the test?
No. It is a breach of the terms, the project gets removed and reviewed by hand, and it is the fastest way to lose good applicants.
What are the diagnostic kinds?
A code snippet, a system design, a piece of copy and positioning, a design critique, or a custom prompt you write yourself. The form names them the way the app stores them.
Can I change the rubric after seeing submissions?
Not to fit a submission you liked. The list exists to be applied to all of them, and changing it afterwards makes the scores meaningless.
Who can see the submissions?
You, while scoring, with the names hidden. The names are revealed once scoring is complete, so the reveal cannot change a score already recorded.
Ready to see it from the other side?
Post a project with a rubric, or browse the freelancers and squads who are already working this way.