Two paragraphs beat nothing, and nothing beats most jobs

Almost every job in this trade is too small to justify a contract, and a minority are too large to go without one. The threshold is not a fee, it is duration and reversibility: how long the commitment runs, and how much you lose if it stops halfway.

The useful version is not a contract in the lawyerly sense. It is two paragraphs in a message, agreed before work starts.

Where the line actually sits

A single job delivered within a week, paid through the platform, needs nothing beyond the two-sentence scope confirmation. The platform's own terms, its escrow and its dispute process are doing the work a contract would do, and they do it better because they are enforced automatically rather than by you.

Write something down when any of these are true.

• The work runs across more than two or three weeks
• You are holding a slot, turning other work away for it
• Payment is staged, or any part sits outside the platform's escrow
• The buyer wants something unusual done with the output - published, forwarded, used commercially
• There is a volume commitment in either direction

Notice that none of those are about size of fee. A large one-off job on normal terms is lower risk than a small recurring one with a vague end date.

Some jurisdictions do draw a fee line in law, and the rules differ by place, so check yours. New York City's Freelance Isn't Free Act, as the city's consumer and worker protection department describes it, requires contracts worth $800 or more to be in writing, counting everything with one hiring party in any 120-day period, and to state the work, the pay and the payment date.

What the two paragraphs cover

Five things, and adding a sixth usually costs you the agreement rather than improving it.

Scope and deliverable. What arrives, in what format, at what length, and explicitly what is not included.

Schedule. Dates for each delivery, and what happens if either side misses one.

Money. Total, staging, and when each stage is payable. If a deposit is involved, say whether it is refundable and under what conditions - the deposit reasoning applies unchanged here.

Revisions. How many, within what window, and what counts as a revision rather than a new job. This is the clause that saves the most hours.

Exit. How either side ends it early and what is owed at that point. Usually: work completed is payable, work not started is not.

That is it. Confidentiality is worth a sentence if the buyer has not raised it, because they generally assume it and are reassured to see it stated. Everything else - liability caps, governing law, indemnities - belongs to jobs at a scale this trade does not usually reach, and inserting them makes a light agreement look like a negotiation.

What platform terms already do

Before you write anything, read the terms you are already operating under.

They typically settle payment mechanics, refund rules, the dispute process, prohibited content, and who owns what. A side agreement cannot override them, and a clause that contradicts them is worse than no clause because it creates an expectation the platform will not honour.

The correct move is to write only what the terms are silent on. That is usually schedule, staging, revision count and exit - the four things a general marketplace has no opinion about because they vary per job. The reasoning behind reading those terms in full, once, rather than at the point of trouble is in the post on what a commission actually buys.

Form matters less than agreement

A signed PDF is not more binding than an accepted message in most places, and it is considerably more likely to stall the booking.

Send the terms in the platform's own messaging, ask for a plain "agreed", and keep the thread. That gives you a timestamped record inside the system that will adjudicate any dispute, which is exactly where you want the evidence to live.

If the buyer wants a document, send one, and keep it to a page. If the buyer sends you one, read every line of the scope and payment sections and ignore the boilerplate; those are the only two places where an unpleasant surprise realistically hides.

When to walk instead

Some requests are a signal rather than a negotiation.

A buyer who will not agree a written scope for a multi-week commitment is telling you the scope is going to move. A buyer who wants payment outside the platform on a large job is removing the only enforcement mechanism you have. Both are cheaper to decline than to manage, and the case for treating declining as a strategy applies most strongly at exactly this size of job.

Buyers commissioning at this scale generally expect the paperwork and are surprised when it is absent - the buyer-side account of how a larger commission is arranged reads as considerably more formal than the average earner assumes. If the job involves committing to a defined scoring structure across many submissions, fix the structure in writing before the first one arrives; the published conventions for consistent scoring are a better reference than a description you invent under time pressure, and the same is true of documented measurement procedure where any of the work turns on a number.

Whether your commitments can be honoured at all in a given week is a scheduling question rather than a legal one, and it is the part earners get wrong most often: agreements are broken by overbooking far more often than by bad faith. Platform-specific terms for judges taking committed work sit on Rate Cock's judges page.