A week of no new components

The rule was simple and slightly stupid: for one week, no new components. Four screens were due. Anything we needed had to be assembled from what already existed, and if it genuinely could not be, we wrote the reason on a list and moved on.

What happened on day one

Day one was miserable. Everyone wanted a component. The settings screen wanted a segmented control, the billing screen wanted a comparison table, and both of those are entirely reasonable things to want. We built them out of a list and a row of buttons respectively, and they looked like the rest of the product for the first time in a year.

What happened by day four

By Thursday the list of genuine gaps had four items on it, and three of them were the same item described differently: we had no way to show a value that was still loading. That is a real hole. We would never have found it under normal conditions, because under normal conditions each of those three screens would have solved it locally and none of us would have noticed the pattern.

What I would keep

Not the rule — a week of it is enough, and a month of it would produce genuinely bad screens. What I would keep is the list. Writing down the moment you wanted something new, without being allowed to build it, turns a hundred private decisions into one shared document. The document is the design system meeting, and it takes ten minutes instead of ninety.

We built the loading component the following Monday. It took two hours, and it went into eleven places.