Improving your proposals is slow. Bidding on fewer bad jobs is instant, and it usually moves your reply rate further in a week than rewriting anything does.
What follows is a filter. It is not a list of clients to be afraid of; it is a list of jobs where the arithmetic is against you.
No budget and a two-line description
This is usually not a cheap client. It is an undecided one, and undecided clients hire slowly or not at all. You are paying to enter a decision that has not started.
The exception: a short post from a client with a long hiring history often means they know exactly what they want and do not feel the need to explain it. Check the history before you skip.
A client who has never hired
No payment verification, no history, no reviews given. Some of these become good long-term clients. Most never hire anyone, and you cannot tell which from the posting.
The exception: a detailed, well-written post from a new client is worth a proposal. The effort in the post is the signal, not the account age.
Forty proposals in the first hour
Past roughly twenty, the client is skimming rather than reading, and your odds of being one of the four they open drop faster than the applicant count rises.
The exception: if you are unambiguously the best match, being forty-first still works. Being a defensible match at forty-first does not.
Nothing specific to open with
If the posting gives you nothing to be specific about, you cannot write an opening that could not have been sent to anyone else, which is the only thing that reliably earns a click.
The exception: none worth the connects. This is the cleanest skip on the list.
A budget far below your rate
Bidding low to get in almost never converts into a higher rate later. The number you start at is the anchor for everything after, with that client and often in your own head.
The exception: a deliberately small first job with a client whose larger work you have seen and want. Decide that in advance, not in the proposal.
Scope that keeps growing in the post
A description that starts as one task and adds three more before the end is describing a project the client has not scoped. That scope creep is already happening and you have not been hired yet.
The exception: quote for the diagnosis rather than the work. A scoped first step is often exactly what an unscoped project needs.
Anything that asks you off-platform immediately
A first message pushing you to another app before any contract exists is the most consistent warning sign on the platform. There is no exception to this one.
Write the rule down
A filter you have to re-decide each time is not a filter. Writing it down is what makes it survive a slow week, which is the only week it is tested.
The arithmetic underneath all of this, including how to work out what you are currently paying per conversation, is in connects and bidding math.