One button swap made a retailer $300 million: how to scope a redesign

The design world loves a rebuild. New look, new demo, big applause in the review. But the case for a redesign has changed. It is less about taste now and more about numbers you already track: what it costs to get a user, whether they trust you, whether they can undo their own mistakes. Let me catch you up on where the argument sits.

When the redesign is really about behavior

A UI redesign is almost always a UX redesign too. Moving a button or rebuilding a checkout changes how people move through a task, not just how it looks. That is why taste alone is a weak reason to ship one.

The Windows 8 case is the warning. In 2012 Microsoft pulled the Start button after seventeen years and dropped in a full-screen tile layout. Experienced users could not find their own apps. The backlash was bad enough that Microsoft put the button back in Windows 8.1 within a year. The lesson for your roadmap: if a redesign breaks the map users already carry in their heads, treat it as a UX decision and plan it like one.

The number your CAC hides

Your cost to acquire a customer counts the ad spend and the sales overhead. It does not count what happens after someone lands and bounces off a bad flow. So a weak UI shows up as rising acquisition cost long before anyone names it a design problem.

The proof is in the checkout. Baymard found 18% of US shoppers abandoned an order because the process was too long, and that fixing checkout usability alone could lift conversion 35%. One retailer forced new buyers to register before purchase. They swapped "Register" for "Continue," made the account optional, and purchases rose 45%, adding $15 million in the first month and $300 million over the year. That is one screen, downstream of every dollar marketing already spent.

Fix the flow, not the whole thing

The instinct to rebuild everything at once is the expensive mistake. Targeted work scoped to the one flow you actually diagnosed earns more trust inside the building and delivers more value outside it. A sweeping overhaul risks breaking flows that were never broken.

So diagnose before you cut. Look for a pattern across signals: support tickets repeating the same navigation complaint quarter after quarter, A/B tests that plateau no matter how many button colors you try, a widening gap between acquisition and retention. A single strong signal backed by data beats three vague hunches combined. When the problem sits in one flow, incremental fixes are lower risk and just as effective.

When you can't explain the decision, they explain it for you

Here is the trap the AI wave is walking into. In 2019 David Heinemeier Hansson posted that his Apple Card gave him a credit limit twenty times his wife's, even though she had the better score and they filed jointly. Wozniak said the same happened to him. Regulators later found no unlawful discrimination. But the product had already failed.

When customers called to ask why, frontline staff fell back on "that's what the algorithm decided." That is a shrug. Explainable AI is booming as an engineering feature and failing as an interface. When a product cannot explain itself, users write their own explanation, and their version is almost always worse than the truth.

Design for the tired, distracted user

Donald Norman's line runs under all of this: humans err continually, and system design should account for it. When something goes wrong, the reflex is to ask who made the mistake. The better question, from Kshama Sinha's read of Norman, is why the design allowed it.

People forget passwords, tap the wrong button, push doors marked pull. Not because they are broken, because they are human. Your job is not a product that never lets people slip. It is one where the slip is safe, understandable, and easy to undo. That standard is also your redesign trigger: if a flow punishes normal human behavior, it is a candidate.

The deep cut

The signals connect, and that is the argument to bring to your next review. A dated UI creates a confusing UX, which shows up as rising acquisition cost and repeat support tickets. If you frame a redesign as a look-and-feel refresh, finance will cut it first. Frame it as recovering the money you already spent to bring users in, and back it with the one flow where the pattern is clearest, and you have a case someone can approve. The move against you is the wholesale overhaul with no diagnosis behind it. Scope to the evidence you actually have.

Three questions for your team

• Which single flow shows a pattern across at least two signals, tickets, plateaued tests, or drop-off, and what would fixing just that flow do to conversion?
• Where in our product does a decision happen that we cannot explain back to the user in plain words, and what do our support staff say when asked?
• Which flow punishes a tired, distracted, first-time user instead of helping them recover, and how much would that cost us in abandoned signups?