I have led twelve significant redesigns. Four of them were good. A couple were actively harmful, and the rest landed somewhere in the middle, which is to say they cost a great deal and changed very little.
Looking back, the failures share a trait. In each case we treated a structural problem as an appearance problem, and appearance is the faster thing to change.
The tell
The tell is almost always in how the project gets described. If the sentence is "the site looks dated", nobody has done the diagnostic work yet. Dated compared to what, and noticed by whom, and standing between them and what outcome?
The good redesigns started from a sentence with a person in it. "People searching for a specific policy give up on the third screen." You can design against that. You cannot design against dated.
Restraint is not minimalism
Restraint is not an aesthetic. Some of the most restrained work I know is dense, colourful and loud. Restraint means every element on the screen has been asked to justify itself, and the ones that could not were removed, no matter who suggested them.
The best thing I ever shipped was a redesign where we changed the navigation, deleted two sections, and left the rest alone.
What I would do differently
Spend the first third of the budget on research and the second third on structure, and only then let anyone open a design file. It feels alarmingly slow. It is faster than doing it twice.
And write down, before you start, the one measurable thing that should be different afterwards. If nobody can name it, the project is a repaint. Repaints are fine, occasionally even necessary, but they should be budgeted and described as what they are.
