Chasing documents is the real bottleneck
A return takes three hours of work and seven weeks of calendar. Nobody bills the seven weeks, nobody schedules them, and they decide how your season goes.

Ask anyone in the practice how long a return takes and you'll get an answer in hours. Ask when the last one went out the door and you'll get an answer in months.
That gap isn't preparation. It's waiting.
Three hours of work, seven weeks of calendar, and almost all of the seven weeks is a file sitting open because one document hasn't turned up. Nobody bills it. Nobody puts it in a schedule. And it decides whether your season ends with a quiet week in June or with everyone in the office on a Saturday in May.
The bottleneck is invisible to every system that measures work
Timesheets record work. Job status records stages. Neither records the thing actually eating the calendar, because "waiting on the client" isn't an activity, so it produces no row anywhere.
Which means the bottleneck can't be managed. It can only be felt. Everyone knows which clients are slow — there's usually one whose name causes a small silence when it comes up in a meeting — but nobody can tell you how much that costs, or whether last year was worse.
The only artefact is somebody's inbox.

First move isn't a new process. It's making the wait visible: for every open file, what's outstanding, who was asked, and when.
"Can you send through your bank statements" is not a request
Most follow-ups don't fail because nobody sent them. They fail because they made the client do the thinking.
Which accounts? Which months? They sent something in March — was that it? A client who isn't sure what you want will put it off until they've got time to work it out, and from your side that looks exactly like a client who's ignoring you.
An actionable ask names three things:
What — in the words the client uses. "Your Westpac business account" beats "the trading account statements", because one of those is a thing they can find in their banking app and the other is accountant.
Which period — "July to September", not "Q1", and definitely not "the missing quarter". They don't know which one is missing. That's rather the point.
Why you need it — one clause, not a paragraph. A client who understands the return can't be finished without it treats it differently from item 23 on a checklist.
Same request, made properly, closes in a fraction of the time. This isn't politeness. It's the difference between one round trip and four, and four round trips is three weeks.

Ask for what's missing, not for everything
The opposite failure is the annual checklist — forty items, sent to everyone, regardless of what you're already holding.
It reads like nobody looked at the file, which is accurate, and it invites the client to skim. Worse, it asks again for things you already have, so the reply comes back full of documents you didn't need with the one you did buried somewhere in the middle.
A chase list should be the difference between what the return needs and what you're holding. Shorter message, faster answer, and it signals the file is being worked rather than processed.
Some gaps are arithmetic and shouldn't wait for a human
If a client lodges quarterly and you're holding three quarters, the fourth is missing. That's subtraction. It should never depend on someone noticing.
Same for monthly statements across a year, or any series where the periods are known in advance. These are the cheapest gaps to catch and the most embarrassing to miss, because the client can see the hole as clearly as you can.
Where the arithmetic is silent — a document with no period on it, an account that may or may not still be open — say nothing rather than guess. A chase list that asks for things which don't exist teaches the client to ignore it, and once that's happened the accurate requests stop working too.

Measure the wait, not the work
If you change one thing this year, record when each document was asked for and when it arrived.
By the end of the season that gives you the two numbers nobody currently has: which clients are genuinely slow, and which of your own requests were vague enough to need sending twice.
Both are fixable. Neither is visible while the only record of the bottleneck is how tired everyone looks in May.