The redesign lands
What I planned on paper was built screen by screen and went through a massive review. The hardest part was deciding what, out of everything, stayed out.
In the previous entry I described studying dozens of apps and drawing up the complete plan for the interface. This week that plan stopped being a document and reached production: the shopping list, search, carts, deals, the weekly planner, the home page. All of it with the same shape on the web and in the app, which was the rule I cared most about holding on to.
Building it was the fast part. What took time was reviewing it. When a redesign touches every screen, each fix can break something somewhere else, and checking screen by screen by hand isn’t enough. I set up a parallel review — over a hundred automated passes, each with its own zone — that found close to fifty details: spacing that touched the edge of a card, a chart that had lost a knot, a button that stayed stuck on the phone after being tapped. None of them were serious. Together they were the difference between a finished product and one that almost is.
I came away with a working rule I didn’t expect: a finding is closed when its reproduction recipe stops failing, not when someone feels it “could be better.” Whatever the review finds beyond that goes on a separate list for the next pass, not back into the same loop.
With the new face in place, I had to look at what sat behind it: what in that interface still rested on data that wasn’t fully trustworthy yet.