The first version of our close screen answered the question “is the period complete?” with a percentage. It was accurate, it was well built, and in eleven user sessions not one person looked at it for more than a second.
What they were actually doing
Every controller did the same thing: scrolled past the summary and started hunting for what was not done. The percentage told them there was work left. It did not tell them what, who had it, or whether it was the kind of thing that blocks the period or the kind that can be accrued.
A completeness number is a fine answer to a question nobody in the middle of a close is asking. They already know it is incomplete. That is why they are on this screen.
The redesign
We replaced the percentage with the exception list and moved completeness to a line of supporting text. Three things changed in the numbers afterwards:
- Time to first action dropped from around forty seconds to under ten.
- The screen stopped being a stop on the way somewhere and started being where the work happened.
- People began arriving at it on the twentieth rather than the first, which is the behaviour that actually shortens a close.
The thing we got wrong twice
Ordering. We first sorted exceptions by amount, on the reasonable theory that the big ones matter most. Controllers re-sorted every time. What they wanted was by blocking-or-not, then by owner — because the useful unit of work is “everything I need from Marta”, not “the largest number on the page”.
Sorting by value optimises for the reader’s importance. Sorting by owner optimises for the number of conversations, and conversations are what the close is actually made of.
What we kept
The percentage is still there, small, at the top. Two of the eleven used it — both CFOs rather than controllers, and both were reporting it upward rather than acting on it. A number that serves a different reader is not clutter; it is just not the headline.
