CRO is useful. I want to say that upfront, because this isn’t an argument against testing. Testing is how you find out what’s actually true versus what you assumed. Run the tests. Pay attention to what wins. That’s good practice.

The problem is when short-term conversion data becomes the only input into design decisions, because it’s measuring one thing while quietly ignoring something else.

A page can be optimized to convert well today and erode trust over time. The two-week A/B test won’t show you that. You’ll see it months later, when customers stop coming back and you’re not sure exactly when or why it started happening. By then the connection to the design decision is invisible.

The most visible version of this is dark patterns. Urgency timers that reset. Pre-checked boxes. Cancellation flows designed to frustrate you into giving up. These things often test well in isolation. They also teach your customers something about you, which is that you’re willing to manipulate them to get what you want. That’s a lesson people carry with them, even when they can’t articulate why they feel vaguely uncomfortable with a brand.

But the subtler version is just as damaging. When every design decision gets made by a metric, you end up with pages that feel engineered rather than considered. The individual choices might all test well. The cumulative effect is a product that reads as cold and transactional, like it was designed to convert you rather than to actually help you. People notice that distinction even when they can’t name it.

The goal isn’t to ignore performance. It’s to hold performance and integrity at the same time. A page that converts well because it’s genuinely clear, honest, and easy to use is a different thing than a page that converts well because it’s built around cognitive shortcuts. The numbers can look identical in the short term. The businesses they build over time rarely do.

↩