Running it
Send it every time, especially when it feels unnecessary
The jobs that go wrong are never the ones you expected to.
Guides on Running it: The admin that turns income into a trade
The argument for confirming scope in writing is settled and covered elsewhere on this site. This is about the part that actually fails: doing it on the job that obviously does not need it.
Selective confirmation protects nothing, because the jobs you would have selected are not the jobs that go wrong.
Why judgement is the wrong tool here
You cannot tell in advance which booking becomes a dispute. The buyer who writes a clear brief and pays immediately is not more or less likely to remember the agreement differently in three weeks; they are just more pleasant to deal with today, and pleasantness is not evidence.
Worse, the moment you start deciding case by case, you have introduced a decision into a workflow that was working. Decisions get skipped when you are tired, busy, or eleven jobs into a batch. A step that always happens costs twenty seconds and never gets skipped, which is the entire reason to make it unconditional.
There is also a signalling effect that runs the other way from what people fear. Earners worry that confirming scope reads as distrust. In practice a buyer reading a crisp restatement of what they are getting reads a professional, and a meaningful share of scope creep never starts because the boundary was stated before anyone thought to test it.
The wording
Two sentences, sent in the platform's own messaging, before you begin.
The first names the deliverable and its size. The second names the date and what is not included.
Confirming: a written assessment across the five axes on my listing, around 600 words, delivered as a PDF. That will be with you by Thursday evening, and it covers the one submission you sent - anything additional is a separate booking.
Adjust the nouns, keep the shape. Deliverable and size, then date and boundary.
Two things must be true of where you send it. It has to be in the platform's own thread, because that is the record a dispute process can read - what evidence actually carries weight is almost entirely a question of where the message lives. And it has to precede the work, because a confirmation sent with the delivery is not a confirmation, it is a description. The date matters beyond this trade's moderators: Stripe's guidance on "product not received" disputes lists evidence that "the agreed-upon delivery date hasn't arrived yet" as a way to answer one, alongside the customer communication.
Making it automatic
Keep three or four versions saved, one per format, and paste the right one. Editing a saved line takes less thought than composing a sentence, and less thought is the objective.
The trigger is booking, not first contact. When payment or acceptance registers, the confirmation goes out before anything else happens, including reading the brief properly.
If the brief turns out to be different from what you confirmed, you send a second one correcting it. Two confirmations on one job is not a mess; it is a record of a scope that changed and was agreed, which is exactly the thing you would want to produce later.
The habit generalises past this trade. Anyone publishing an assessment somebody else relies on ends up stating conditions before producing a figure, which is why the measurement side documents its method up front rather than afterwards. On the buyer's side the same message does quiet work: the account of how a commission normally opens treats a clear restatement as the point at which a buyer stops worrying, which is why the confirmation converts as often as it protects. And it is the difference between a person and an interface - the automated scoring tools confirm nothing and agree to nothing, which is a fair part of why buyers who want a commitment come looking for someone who can make one. For the platform-side view of where these threads live and what is retained, Rate Cock's judges page is the current source.