The science of A/B testing.
A/B testing is a proven means to increase conversion rate. It's not as simple as it appears. It's more than simply testing two or more versions of Web pages, banners, search ads, or whatever persuasive element you can imagine. It also has limitations.
A/B testing, unlike the intuitive creative process, must be treated like a true scientific experiment. Many clients come to us frustrated by the results of their tests. Almost without fail, we discover their tests don't conform to experiment design. Most just heave two versions of an element at the wall to see what sticks.
Unless you're one of those learn-the-hard-way types, benefit from our experience. Here's what we've learned from seven years of trial and error.
One Change at a Time
When beginning an A/B test, establish a control and baseline for whatever you're testing (e.g., Web page, landing page, banner ad). Put your best creative effort online first. Monitor its performance over an established period.
Once a baseline is established, start optimizing the page by making one change at a time. More than one makes it impossible to determine which change actually made a difference.
A client wanted to change the way the box for their lead-generation form appeared. They thought they were testing just the form box. I pointed out changing the background color, headline, copy, and box position amounted to multiple variables and should therefore be conducted over multiple tests.
This is a limitation of A/B testing; conversion improvements can only be made incrementally.
Clicks Don't Exist in a Vacuum
Customers participate in scenarios on your site or in your marketing efforts whether you plan them or not. Each page or touch point exists in the context of what visitors saw on a previous page. When conducting an A/B test, consider all scenario elements. Often, a change to another element affects test page results.
A page with high exits or a call-to-action form with high abandonment may not actually be problems. Consider the page visitors viewed prior to the test page. How many of us have clicked a "Free iPod" banner, only to find a landing page detailing the painful process of getting a free iPod? We could optimize and A/B test the landing page ad infinitum, but the real problem is the misleading banner.
Recently, to increase an email open rate, we tested a new "free offer" subject line. The email body remained the same. Though the open rate went up, conversion rate and total conversions dropped. Instead of returning to the original subject line, we modified the email copy. Test copy conversion strongly outperformed the original.
The new subject line changed the email's context. Until we changed the body copy to echo the subject line, we were losing momentum. Users in the first test of the new subject line were simply not seeing the free offer in the original body copy.
In complex selling scenarios, many sites successfully funnel high amounts of traffic to a call-to-action form. Yet they test different versions of the form with little success. Usually this is indicates prospects don't have enough information or resolve to complete the form. The problem isn't the form; it's the scenario visitors participate in.
Small Changes, Big Differences
Never assume a seemingly inconspicuous change will bring inconspicuous results. The day of the week, a font, or a color palette shift can dramatically affect test results. Just moving a few buttons around on a checkout page can mean deep decreases in conversions.
There are really only 20 design principles that can affect conversion. Don't take any of them lightly.
Testing Two Very Different Creatives
Many conduct an A/B test on two significantly different offers or two significantly different creative efforts.
If effort A returns a conversion rate of 3.0 and effort B returns a conversion rate of 2.5, many will abandon effort B for the seemingly more efficient A. This could be a colossal mistake.
That's not a true A/B test. There's no way to determine if effort B were optimized if it perform better than optimized A. Think of this as two different tests rather than as one.
Often, we've found the two efforts speak effectively to two very different audience segments. Both may be worthy of further optimization, and their cumulative conversions would go well beyond effort A alone.
A/B Testing Is Real Work
True, effective A/B test is not for the impatient. You must be committed to true scientific application. Next time you move into the A/B laboratory, remember that A/B testing is science, not art.
Meet Bryan at Search Engine Strategies in Toronto, Canada, May 4-5, 2005.
Bryan Eisenberg is co-founder and chief marketing officer (CMO) of IdealSpot. He is co-author of the Wall Street Journal, Amazon, BusinessWeek, and New York Times best-selling books Call to Action, Waiting For Your Cat to Bark?, and Always Be Testing, and Buyer Legends. Bryan is a keynote speaker and has keynoted conferences globally such as Gultaggen, Shop.org, Direct Marketing Association, MarketingSherpa, Econsultancy, Webcom, the Canadian Marketing Association, and others for the past 10 years. Bryan was named a winner of the Marketing Edge's Rising Stars Awards, recognized by eConsultancy members as one of the top 10 User Experience Gurus, selected as one of the inaugural iMedia Top 25 Marketers, and has been recognized as most influential in PPC, Social Selling, OmniChannel Retail. Bryan serves as an advisory board member of several venture capital backed companies such as Sightly, UserTesting, Monetate, ChatID, Nomi, and BazaarVoice. He works with his co-author and brother Jeffrey Eisenberg. You can find them at BryanEisenberg.com.
2015 Holiday Email Guide
The holidays are just around the corner. Download this whitepaper to find out how to create successful holiday email campaigns that drive engagement and revenue.
Three Ways to Make Your Big Data More Valuable
Big data holds a lot of promise for marketers, but are marketers ready to make the most of it to drive better business decisions and improve ROI? This study looks at the hidden challenges modern marketers face when trying to put big data to use.
December 2, 2015
1pm ET/ 10am PT
Wednesday, December 9, 2015
5pm HKT / 5am ET