That’s right, I want you to fail.
I want you to fail yesterday! It is through failing that your website, your next email, your next landing page will achieve its ultimate potential. But there’s one catch to this plan: In order to fail your way into success, you must test.
Why the focus on fail?
We will spend the rest of this article talking about various methods of testing, and when marketers begin this endeavor, they must be hungry for failure. Take, for instance, a multivariate test with 128 variations. What I really hope to see out of this test is 127 failures that will lead me to the one rock-star combination that provides a 187% lift in conversions. That’s what I want, and I want it fast.
So, do I really want your brand to fail?
Well no, of course not. the evolution of brands online has brought a much-needed emphasis on analytics, something the c-level executive is now well aware of. But just knowing where you stand and how well something performs currently against no control means you know nothing more than, well, where you stand.
What is the one thing most of us want to do as marketers?
Kill the competition! While a recent Jupiter research/ e-rewards study found that only 24% of marketers are currently testing, it also showed that 30% planned to on-board testing activities within the next 12 months. If you are not planning on testing in the near future, you can bet your competition is.
Want to find out how to move beyond analytics and into action? Then, read on. Here’s what we’ll cover: A/B / Multivariate testing Tips+Tools
A/B testing is hardly a new concept. Well employed by direct response/direct mail marketers for many years, it is the online permutation we will be discussing in this article. in a/b, as well as multivariate testing, we are testing for conversions. But unlike multivariate testing that allows for simultaneous tests of many elements, a/b tests focus on a single variable. Think in terms of testing two versions of the same web page. One version would have an image on the left, text on the right, and a call-to-action button on the bottom. Then, imagine reversing the position of the image and text and voila!—you have a second version ready for an a/b test. Want to add more versions? sure thing—just make it into an a/b/c/d test; the concept remains the same. It is for this reason that many call this a Bucket Test.
Multivariate testing allows for real-time testing of many elements running simultaneously. This testing strategy is attractive to marketers because it allows for hundreds (if not thousands) of recipes or variations to be tested all at once using statistical modeling. In this scenario, think of a web page with a headline, a benefits paragraph, a call-toaction, an image, etc., and then imagine wanting to test five variations of each. Multivariate tests were made for this.
To make it simple, this type of testing allows marketers to determine which combination of elements performs the best, and it allows them to find this information quickly.
Tips for Testing
1) Keep it simple! often the first mistake made in any testing endeavor is the propensity to want it all. We want to test everything under the sun and can’t decide on a single element, so we target 14 different elements on a page. Stop! the first thing to remember is simplicity. start with a simple a/b test.
2) Choose only one goal or Conversion. Whether you are testing email subject lines or landing pages, the key is having a singular conversion. You can even think outside the box to focus on increasing the user’s time on the page or clicks to purchase. Perhaps you want to increase downloads. Whatever your approach, make sure your target doesn’t become targets (plural)!
3) Pick the lowest hanging Fruit First. In other words, test the more prominent elements first, like the number of columns or text vs. image. These pieces can have a large impact on conversion lift. Save the color of buttons and type of font for last. These will not have as big of an impact even though they are still worth testing down the road.
4) Test out high-traffic Pages First. This way, if you have less popular pages with similar users, you can simply apply the winning layout to these pages. Remember, it is important to only take this shortcut if you are sure the users on each page are the same. Do not assume the tests for a page about 18th century violin concertos would yield the same user and result as a page about 21st century rock-n-roll ballads.
5) Always Run tests in a split Fashion when testing a/b. Testing an a version for one period of time against a b version during another period of time enters unnecessary variables into the mix.
6) Don’t assume you know how your consumers will act based on new variables. testing is the only way to know the true answers— sometimes validating your expectations and often opening your marketing eyes to a whole new world of conversion.
7) Protect your Brand! Testing can lead to decisions that will diminish your brand in exchange for higher conversions: “I wonder if my logo should be much smaller? Or, do I even need my logo?” Sounds crazy, right? Yet, I know plenty of colleagues who make these mistakes daily. Remember that making a choice to diminish your brand on one page can have devastating effects across the many other pages or transactions consumers will have with your website. In a world where many brand journeys start and end online, choosing to diminish a brand is dangerous. Testing is not an excuse to turn your online brand experience into giant red buttons and blue hyperlinks.
There are many tools out there with varying price ranges, but let’s take a look at two solutions on each side of the spectrum. What separates these two very distinctly is both price and multivariate methods.
Google Website Optimizer google provides a well-designed and free solution that allows for enterprise-level testing on landing pages and web pages. While I would recommend seeking out the assistance of a google-authorized agency, this system can yield significant results and has a price that is hard to beat.
Google’s multivariate technique employs a full-factorial method for testing. This means that every possible combination is served while testing. While some analysts swear by this testing method, it can take longer to perform tests on areas with low traffic.
Omniture Test&Target a true enterprise solution, Test&Target includes some options that google does not, such as email testing that allows for serving winning images on the fly. The cost is nowhere close to free, but training and consulting pay off long-term.
Omniture’s multivariate testing strategy uses a fractional-factorial methodology—called the Taguchi Method— that allows for taking samples of smaller traffic numbers before using algorithms to create answers and predict winners.
So Let’s Fail Now!
Remember how we opened this article? The concept of failing your way to success is strong! this is the time to begin testing your marketing online and to create failures in order to achieve success. The preconceived notion that you don’t need testing and that you are fully aware of what your consumers want online is dangerous and presents unnecessary risk. And if you feel hesitant to agree with that, I challenge you to prove me wrong.