The proposal is the first piece of an agency's actual work you will ever see. Everything before it was conversation.
So read it the way you would read a deliverable, because it is one.
A restatement of the problem, in your words. Before any solution. If the opening pages are about the agency, nobody listened hard enough on the call. You should recognise your own business in the first page, including the parts you find slightly embarrassing.
A scope with edges. What is included, and explicitly what is not. Exclusions are the single most reliable sign of experience in a proposal. They exist because someone got burned on that exact item before.
The assumptions it depends on. Access, approval speed, someone available on your side, the state of your data. Every fixed price rests on assumptions, and the honest ones write them down.
A sequence, with dependencies. Not just phases with dates. Which things must happen before which other things, and what happens to the timeline when one slips.
What you own at the end. Accounts, documentation, configuration, creative files. This should not require asking.
How change is handled. Something will change. There should be a stated mechanism for it, priced or at least described.
No exclusions means either nobody has done this work before, or they intend to negotiate scope later, when you have less leverage.
No assumptions means the same risk, held on their side, which sounds good until it becomes a reason the timeline moved.
No client-side effort described usually means it was not estimated. It does not mean there is none.
No handover terms is worth asking about directly. Sometimes it is an oversight. Sometimes it is a retention strategy.
A timeline with no dependencies is a wish. Real projects have things that cannot start until something else finishes, and a plan that does not show them has not been thought through.
A sixty-page proposal is often a template with your name in it. The tell is how much of it could be sent to any other company in your industry without editing.
Scan for specifics. Your systems named. Your actual numbers. The constraint you mentioned on the call, reflected back with a response to it. A short proposal full of those beats a long one full of methodology.
The proposal that arrives within a day of the first call, fully priced, for a complex build.
Either it is a template, or the work is more standardised than they implied, or nobody looked at your systems before quoting. All three are useful to know, and all three mean the number is a guess that you will be renegotiating later.
There is a version of fast that is fine: a clearly-labelled estimate or a range, offered as a starting point. That is honest. A precise number produced without discovery is not fast, it is unfounded.
You do not need to grade proposals on every point. Read for two things: does this describe my business specifically, and does it admit anything.
A document that is specific and admits limits is written by someone who expects to be held to it.