Shopify Store Design Mistakes That Lower Sales (and How to Redesign Safely)

Here is a five-star review that cut my sales by 62%. It’s a real review from a real customer. It reads beautifully, and on paper it’s exactly what you’d want a shopper to see just before they check out. When I A/B tested it on my own e-commerce store, orders fell from 13 to 5.

One test, one change, one clear answer. Now picture the same review as one of thirty changes bundled into a brand-new website. You launch the redesign, sales drop, and you have no way of knowing which of the thirty changes caused it. That is what happens to most redesigns, and it happens to some of the biggest retailers in the world.

In this article I’ll walk you through two famous redesign disasters, what Microsoft learned from thousands of experiments, two beautiful tests from my own store that lost, the tests that actually won, and the five-step process I use so a redesign makes money instead of quietly losing it.

The pattern I’ve seen in 100+ brands

I’ve worked in digital marketing for more than 15 years. I was head of growth at Karma and Luck, a nine-figure brand, where we grew the conversion rate by 50% in one year of non-stop A/B testing, and I’ve worked with more than 100 e-commerce brands.

Over those years I’ve watched the same story repeat itself. A brand spends tens of thousands of dollars and months of work on a new website. It looks amazing and the whole team loves it. They launch it, and the conversion rate goes down. Then everyone blames the ads, the season or the economy: anything except the new website.

Beautiful doesn’t mean it sells. Ask yourself: when did you last change something on your store because it looked better (a banner, a font, a new product page) and then measure what it did to sales? Most store owners never do.

Redesign disaster #1: Finish Line lost about $3 million in three weeks

Finish Line, a large American sneaker retailer, launched a brand-new e-commerce website right before Black Friday. Customers complained, things broke, and online sales came in below plan. Three weeks later, on December 6th, they switched back to the old website.

Those three weeks cost them about $3 million in revenue. Their CEO admitted that launching the new site in November had been a huge mistake, and their chief digital officer left the company soon after.

The new site wasn’t ugly. The problem was that nobody knew how it would perform until it was already live for every customer, at the worst possible moment of the year.

Redesign disaster #2: Marks & Spencer and one hidden change

Marks & Spencer relaunched its website after years of work and an investment of around £150 million. Buried inside the project was one change that sounded small: every customer had to register again.

Only part of their customers came back and did it. Registrations dropped, conversion dropped, and online sales fell 8.1% in 13 weeks. One small change hidden inside a huge redesign, exactly like my review in the cart.

Why smart teams with big budgets get it wrong

Because nobody can reliably predict what will increase sales. Not designers, not agencies, and not me.

Microsoft has run thousands of A/B tests. Every one of them started as an idea that someone smart believed would improve the business. What they found:

Outcome of the idea	Share of experiments
Made things better	about 1/3
Made no difference	about 1/3
Made things worse	about 1/3

On Bing, their search engine, about 80% of experiments were flat or negative. These are some of the best product teams in the world, with enormous amounts of data. If their ideas fail two times out of three, what are the odds that every idea inside your redesign is a winner?

There’s even a name for how most website decisions get made: the HiPPO, the highest-paid person’s opinion. The founder likes the new design, the agency likes it, so it goes live. Nobody asks the only people whose opinion really counts: your customers. And the only way to ask them is a test.

What happens when you launch 30 changes at once

Let’s say your redesign contains 30 changes: a new header, new product page, new cart, new colours, new copy, new photos. If it behaves like Microsoft’s data, roughly 10 of those changes help, 10 do nothing and 10 hurt your sales.

Launch them all together and the good ones and the bad ones cancel each other out. Sometimes you end up flat after months of work and tens of thousands of dollars. Sometimes you end up worse. Either way, you’ll never know which change did what. That’s why everything you believe will increase sales has to be A/B tested. Everything.

Two beautiful tests that lost on my store

Test 1: a five-star review in the cart drawer (−62%)

The idea was simple. People are about to check out, so show them a happy customer: more trust, more orders. It looked beautiful.

Same store, same traffic, about 5,800 visitors on each side. 13 orders without the review, 5 with it. Minus 62%. I stopped the test.

Why would a good review hurt? In the cart, people have already decided. They don’t need more convincing. A review there is one more thing to read right before checkout. And this particular review mentioned that the team sent photos and a video when the customer asked. That planted a brand-new doubt: should I ask for photos first? So people left to ask, and many never came back.

If that review had been part of a redesign, it would still be live today, quietly hurting sales every single day. (The same idea placed before the decision, under the Add to Cart button, added $67,365 on another store. Placement is everything.)

Test 2: an illustrated style guide (losing so far)

I genuinely like this one. An illustrated style guide on the product page: four ways to use the product, and what each one says about you. It’s honestly the most beautiful thing on the page.

Seven orders without it, six with it. Right now it’s losing. I’ll let it run a little longer, and if it still loses, it goes, even though I love it.

The tests that won

Now look at what did work on the same store:

Test	Result	The doubt it answered
A trust box with four simple lines: certificate, insured delivery, returns, secure payments	+17% conversion rate	Can I trust this store?
A size guide that shows people how to measure	+36% (10 → 14 orders)	Will it fit?
A new homepage with four products and one button each	+38%, still running	Where do I start?

The size guide is a good example. The size was already written in the description. People could read it, but they still didn’t know whether it would fit them. On a $1,000 product, the question isn’t “do I like it?”, it’s “will it fit?”. So we showed them how to measure, step by step, right on the page.

None of the winners was the most beautiful idea. They were the ones that answered a real doubt: Will it fit? Can I trust this store? Can I return it? I go deeper into each of these in six real tests from a high-ticket store.

Does that mean you should never redesign?

No. It means you redesign the right way. Here are the five steps I use.

Step 1: write down every change

List every single change in the new design: new product page layout, new reviews section, new menu, new checkout button. Most redesigns are 30 to 50 changes pretending to be one project.

Step 2: rank the changes by the doubt they solve

What stops people from buying on your store right now? Will it fit? Is it worth the price? When will it arrive? The changes that answer the biggest doubt go first. The ones that only look nicer go last.

The biggest doubt depends on what you sell:

• Fashion: will it fit?
• Skincare: will it work for my skin?
• Furniture: delivery and returns.
• Electronics: will it work with what I already have?

Find that doubt and you’ve found your first test.

Step 3: test the changes on your current website

Test them one by one, or in small groups. Half your visitors see the old version, half see the new one. Wait for about 50 orders on each side before you decide. And if you sell expensive products, judge the test on conversion rate, not revenue: one big order can make a losing version look like a winner.

Step 4: keep the winners, drop the losers

Your new website is simply all the winners together. No guessing, no hoping.

Step 5: if you must relaunch, don’t switch everyone at once

Sometimes a whole new site is unavoidable, for example when you move platforms. In that case, send part of your traffic to the new site first and compare it with the old one. Never launch right before your biggest season, as Finish Line did. And never make customers start over, as Marks & Spencer did when everyone had to register again.

“Testing every change sounds slow and expensive”

With an agency, it is. A CRO agency can cost $5,000 to $10,000 a month, and you wait weeks for a single test.

That isn’t how I work anymore. I use AI to find what stops people from buying, build the new version in the same style as the store, and launch the test in hours. No developers, no designers. Every test in this article was built that way. (Here’s how the AI CRO process works.)

So before you spend $20,000, $30,000 or $50,000 on a redesign, spend a few weeks testing the changes inside it. You’ll keep the ones that make money and avoid the ones that would have quietly killed your sales.