Forty pages, PDF, a three-level table of contents, screenshots for every feature point, even 4px corner radii annotated on buttons. Clearly written with care.
We didn't quote it as-is. We booked a half-day session, copied all 23 feature points onto a whiteboard one by one, and asked three questions about each: who uses this today, what happens if we don't build it, and can it already be done somewhere else.
We crossed out 11.
Six were "we might use this someday"; asked roughly when, the answer was basically "hard to say". Three could be done with the Excel they already have — filtering a list by month is a two-minute pivot table. Two more duplicated features in the client's other system; neither side knew the other had it.
The remaining 12 went into two phases. Phase one took 5 — all of them "people use this daily right now, and without it they're copying data by hand".
Halfway through the meeting the doc's author got a bit anxious: we'll need all of this eventually. We photographed the whiteboard and sent it to him: the doc stays, phase two follows it — just don't let "eventually" take the seat of "right now".
Three weeks after phase one launched, the client messaged us: 4 items from phase two were no longer wanted, because once people settled into the new flow, the process had changed. Those 4 had been near the crossed-out list from the start.
That 40-page PDF still lives in the project folder, filename untouched. We occasionally open it — mostly to admire the TOC. It is very tidy.
