Design Optimization: Are You Still Not Split Testing?

Many beginning designers share a common misconception that design is all about creativity and inspiration, and of course the necessary technical skills. Certainly these are essential, but professionals know that the bottom line when it comes to design is performance. If you consider yourself more an artist than a designer, this is never an issue. But if you are doing design work for a client, it’s THE issue. You may create something very elegant and original, but in almost all cases the real goal is not just improved appearance but improved outcomes.  And to know if your design is really working as intended, or could be improved further, you have to test it.

In other fields the need for testing is a given. You wouldn’t design and market a physical product without ever doing any basic testing, and in a competitive environment you might test, rework and retest your product repeatedly. In the online world, savvy internet marketers constantly test variations of their offers and landing pages to determine which result in the highest number of conversions. But the benefits of split testing are not restricted to internet marketing: visual website optimization can be important in almost all types of web design.

Basic split-testing entails dividing the traffic to a given web page between two alternate versions and seeing which produces the best response, in terms of some specific, measureable goal. In simple A/B testing, the two pages may be identical except for a single variable: alternate header logos, different headlines, different placement of a button or image, etc. Or two entirely distinct versions of a home page or landing page can be tested, as may be desirable when comparing an existing versus alternate design. In multivariate testing different combinations of three or more variants can be tested simultaneously, though this requires a larger number of site visits to give meaningful results.

Conversion Optimization

The performance measure being tested for depends on the type of site. For any sort of commercial website, obviously, the key variable might be the rate of sales conversions. In other contexts a “conversion” might be having the user click through to a particular page, fill out a form, or provide an email address. Even for sites that are entirely non-commercial and minimally interactive, you might want to test how different logos, layouts or CSS styles affect things like the time users spend on the site, or the bounce rate. Anything that can be tracked in analytics can compared and optimized using split testing.

The prerequisite for this kind of audience response testing is, of course, an audience. Split testing works best for websites that are already receiving traffic, and is ideal for trying out a new look and/or refining a few critical elements as part of a design change. The more traffic a site receives, the more meaningful results can be generated in a shorter amount of time, because the sample size is larger. If Alternative B gets 7 conversions and Alternative A gets three, that might indicate B is outperforming A or it might be due to chance. If instead the numbers are 700 and 300, B is clearly a winner. There are ways to test the statistical significance for these kind of test results, when it’s important to be able to draw firm conclusions. But even partial test results can be illuminating–probably more so than the feedback you may get from just a few individuals offering their opinions about the effectiveness of your work.

Adaptive Design

Split testing can also be used when you are creating an entirely new site. You can set up tests on alternate site designs from the outset, which will generate useful results as traffic starts to build, eventually leading to improvements in performance. This sort of “adaptive design” approach might potentially add a new dimension to your practice, since rather than being hired for a single one-time project you could provide an ongoing service of website design optimization. Alternatively, you could deliver a simple A/B testing setup to your client as an added feature along with your design package, using a simple and easy-to-use split testing product such as Visual Website Optimizer. The client would then be free to experiment with different heading text, opening paragraphs, etc., without ever having to insert code or modify site files at all.

In some situations it might be appropriate to split test design options by buying traffic, using pay per click advertising. This is an option worth discussing with your client, even if they have no intention of running a pay-per-click or pay-per-view campaign on a regular basis. For any kind of commercial, business or professional site, an initial investment in traffic generation to conduct testing, with an aim toward optimizing conversion rate, could yield significant benefits.

If you are designing a website for a client, the advantages of testing to optimize conversions is something you may need to explain. Most people are at least somewhat aware of search engine optimization, and will expect or request that SEO considerations be taken into account for their new site. Often however both designers and their clients pay less attention to how different design variations may affect conversion rate–which is strange because for any site in which some kind of response from the viewer is being sought, conversion rate is the critical multiplier. A site that converts at two percent does twice the business–or enjoys the benefits of twice the traffic–as a site that converts at one percent.  Something as simple as a differently sized logo or re-placement of an email opt-in form might be all it takes to get that one extra visitor in a hundred to convert, but there’s no way to know until you try.

Getting Started

If you have never done any split testing, a great site to learn more about it is ABTests.com. Here you can see many examples of tests people have actually run.  Each example includes a write-up with images of the two pages being compared, an explanation of exactly how they differ, and test results showing which was the winner and discussion of why one may have favored over the other.

A number of products and services are now available to help make split testing easier and more efficient. There’s also , which is free, but costs more time to use and manage well. In a future post I’ll write more about the tool I am now testing out, Visual Website Optimizer. I’m only a few days into the free trial so far, but my first impressions are that it’s extraordinary. At least in terms of setup and ease of use, I haven’t seen anything else like it. You can sign up for a free 30 day trial, set up your first test in 30 minutes, and start getting results immediately. I’ve set up an affiliate link here because I think it’s a brilliant tool. But whether you click through from here or not, check them out at http://visualwebsiteoptimizer.com, and consider using the free trial to start running some A/B tests on some of your own sites.

Tagged as: testing, Web Design