Before you commission a redesign, run these four weeks of research

Why redesigns disappoint

A redesign is a very expensive way to change many variables at once. When conversion goes up afterwards, nobody knows which change did it. When it goes down — which happens more often than agencies admit — nobody knows either.

The uncomfortable truth is that most sites that "need a redesign" have three or four specific problems and a lot of pages that are basically fine.

Week one: the numbers

Start in analytics, not Figma. Build a funnel by device and by traffic source. You are looking for the step where the drop is disproportionate, not the step where the drop is biggest — those are different things.

Segment by new versus returning. A store that converts returning visitors well and new visitors badly has a trust problem, not a usability one, and no amount of visual polish fixes it.

Week two: the recordings

Watch forty sessions. Not four hundred — forty, properly, with a notebook. Watch the ones that reached checkout and abandoned, and the ones that bounced from a product page after more than thirty seconds.

You are looking for hesitation: repeated scrolling over the same block, opening and closing the same tab, rage clicks on things that are not buttons.

Week three: ask customers

Put a two-question survey on the product page and one on the order confirmation. "What almost stopped you buying today?" on the confirmation page is the single highest-value question in eCommerce research.

Then call five customers. Twenty minutes each. You will hear the same sentence three times, and that sentence is usually your biggest fix.

Week four: the decision

Now you can answer honestly: is this a set of specific fixes, or is the underlying structure wrong? Both are legitimate outcomes.

In our own client work, roughly two in three "we need a redesign" briefs turn into a testing programme instead. The other third genuinely do need rebuilding — and those projects go far better, because we know what to fix before we start.

If any of this is a live problem rather than an interesting read, say so. We would rather have a specific conversation than a general one.

— Sayan Saha, Founder