The Threats of a Radical Site Redesign

Neil McKay

It is easy to assume redesigning your website will lift conversions and eliminate problems. But, a radical redesign is costly in terms of time and money—and you could be taking a big risk. The scope of the project can change, adding costs and creating pressure as the launch date, which is often just before peak selling season, approaches. More often than not, therefore, the lift the company was expecting from its radical redesign isn’t achieved.

This happens for a number of reasons:

  • Existing repeat customers are confused by your new layout. Things are not where they used to be and, as a result of this confusion they don’t spend their money. Instead, they leave and you lose revenue.
  • Your new design may have fixed some of your conversion problems but it has also created others. These new problems either cancel out your lifts or bring down your average conversion figure. This is difficult to predict, because you leapt into the unknown, basing a fair proportion of your design decisions on gut feeling and intuition, rather than on data.
  • You didn’t need a radical redesign in the first place. You made the call too early. You thought you had exhausted all opportunities within your existing site and had reached your local maxima, but you hadn’t.

The false positive of a redesign

Even if you do see a positive conversion lift, the chances are you have still lost out. You have lost out on valuable “learning” which, for conversion optimisation, is crucial!

You must think about all the changes you made in the redesign—the layout, navigation, content, images, video, page flow, and call to action invitations. How do you know, conclusively, which design changes were responsible for your conversion change (positive or negative)?

You don’t know, because you did not test each change individually—in isolation and within a controlled test environment.

In his book ‘You Should Test That!‘, Chris Goward explains the importance of targeted testing and shows how website redesigns can mask the effects of the many individual changes.

So how do you minimise these radical redesign threats?

Essentially, I should advise you to avoid a radical redesign whenever possible. Instead, evolve your design carefully. Make small changes and test them to find your conversion improvements. And keep on tweaking and testing until you reach the point of diminishing returns.

Clearly, the fact that you are looking at a redesign suggests that you feel you have already reached this local maxima. At this point, stop and go back to conducting conversion research across all relevant disciplines: experienced-based assessment, and qualitative and quantitative research.

By gathering important data about your current site (which you may not have done for a while), you will gain new insights that help you to understand:

  • Who is using your website?
  • How they are using it?
  • Why are they there? What are they trying to achieve?
  • What problems are they having in achieving their objective?
  • What is distracting them?
  • What annoys them about your website?
  • Are they motivated and able to achieve their objective?
  • Are you making it easy for them?
  • Are you encouraging them to act?
  • Are they doing what you want them to do?

This level of insight will help you decide if a new radical redesign is, in fact, needed. If it is, you can go ahead on the strength of fresh knowledge and insight.

Indeed, the data will help direct your new radical redesign, allowing you to keep the things that you know work well and to start from scratch with the things that don’t. And, importantly, you will have more confidence about where your time and budget needs spending.

The balance between your data-backed methodology and design intuitiveness and heuristics will be better, allowing you to make better, innovative, and optimised design decisions. This newfound insight greatly reduces the risk in redesigning, and has improved your chances of conversions whereas before you were tempting fate by acting on instinct rather than on robust data.