Case study · 2026
Grouped: A Live Product Re-skin in Under Two Months
Seven weeks to reskin a four-year-old product without pausing it. The sequence outlived the pixels.
01
The problem
The product looked like three products. Four years of shipping had left two button systems, three grays that all claimed to be the same gray, and a navigation pattern that changed depending on which year a screen was built in.
Nobody disagreed that it needed redoing. The disagreement was about stopping. Engineering wanted a freeze and a clean rebuild. Support wanted nothing to move at all, because every visual change generates tickets. Both were right about their own cost and wrong about the other one's.
02
The strategy
We stopped treating it as a redesign and started treating it as a release sequence.
The order was decided by usage, not by how bad a screen looked. The screens people open daily went first, not because they were worst, but because that is where a new system either earns trust or doesn't, and we wanted to know inside two weeks rather than at the end.
Half the work was getting product, design and support to agree on one word for the thing we were shipping.
Each release carried its own changelog line, written before the work started. A release whose changelog needs two sentences is two releases.
03
The execution
Eleven releases, each one a complete thought. Every release had a one-line changelog written before the work started, which is a cheap way of finding out whether a release is actually one thing.
Support got the screens a week ahead with a short note on what would generate tickets and what to say. That week of lead time cost nothing and removed the loudest objection to shipping at all.
04
The outcomes
Seven weeks, start to finish, with normal feature work continuing alongside. Support ticket volume moved inside its usual weekly range.
The more useful outcome was the sequence itself. Grouped now has a default shape for shipping anything visual that touches every screen, which is a thing you can only get by having done it once under real conditions.
05
What went sideways
Release six broke the thing we had been most confident about.
The new navigation tested well in isolation and badly in sequence. Someone who had already learned two of the three old patterns on the way through a task now had a third to learn, and we had sequenced by screen rather than by journey, which is a real distinction we had written down and then not used.
So we stopped mid-run and shipped a bridge release that did nothing except make the old and the new navigation behave identically for a fortnight. Nine days, no features, no visible progress.
The plan survived. The confidence did not, and that was the useful part.
06
What it taught
Sequence. A freeze is a confession that you do not know what order to do the work in. Decide the order by what people open rather than by what looks worst, and the freeze stops being necessary.
Warn the warners. The team that handles the complaints should see the change before the people who will complain. A week of lead time cost nothing and removed the loudest objection to shipping at all.
One word. Product, design and support have to agree on what they are shipping before they can agree on when. "Re-skin" won because it promised less than "redesign", and we could keep that promise.
Credits
- Design lead Grouped product team
