Every few weeks someone asks us to redesign a product that does not need redesigning. The instinct is understandable. The product feels tired, the metrics have flattened, a competitor launched something sharper, and a redesign is the one intervention big enough to match the feeling. It is also the one intervention that puts everything at risk at once.
The honest question is not whether the product looks dated. It is whether the model underneath it is wrong. Those are different problems with different price tags.
The test: is the structure wrong, or the surface?
Refinement fixes the surface. Wording, hierarchy, spacing, states, the order of two steps, the thing that is three taps deep and should be one. It is cheap, it ships continuously, and it compounds.
A redesign is warranted when the structure is wrong: when the information architecture no longer matches how people use the product, when the business has changed shape, when three years of features have been bolted onto a frame built for one. No amount of surface work fixes a wrong frame.
If you cannot name the structural thing that is broken, you are asking for a redesign to avoid making a decision.
Four questions we ask first
- Where do people actually fail? Not where you suspect, but where session recordings, support tickets and five usability sessions agree.
- Has the audience changed? A product built for early adopters and now sold to enterprises has a structural problem, not a visual one.
- How many of your complaints are about copy? If most of them are, you have a writing project wearing a design costume.
- What did the last redesign change, and did it work? If nobody measured, the next one will not be measurable either.
What refinement looks like when it is done well
A refinement programme is not a coat of paint. It is a queue, ordered by evidence: the twelve smallest changes that remove the most friction, shipped in sequence, each one measured. On one fintech engagement, the first four items in that queue moved activation more than the redesign the team had budgeted for, and cost a fraction of it.
It also buys you the thing a redesign cannot: knowledge. Six weeks of shipping small changes tells you exactly where the structure resists you. That is the brief for a redesign, if you still need one.
When we do recommend the full rebuild
Three signals, and we want at least two of them. The navigation has more than one home for the same task. The team cannot add a feature without arguing about where it goes. And the interface has stopped being able to say no to anything, so every screen carries every option.
When those hold, refinement is deck-chair work. Rebuild the model, then let refinement do what it is good at.
The cheapest version of the right answer
Spend two weeks before you spend two quarters. Audit the flows, sit with five users, list the failures, then split them into surface and structure. The split is the decision. It is almost always cheaper than the meeting where everyone argues from taste.