Split Traffic, Raise Conversion Rates, Repeat, Multiply

Site optimization is the science (and art) of tweaking the pages of your site to increase conversions. Although it’s one of the most under-appreciated disciplines of Web marketing, site optimization doesn’t land on the priority list of most companies. In fact, few companies optimize their sites at all. It’s a shame because there’s so much for a company to gain: for example, doubling revenue.

Think of it this way. What are the chances anyone has ever perfectly designed a Web site on the first try? The answer is zero percent. It’s just not possible to come up with the optimum solution through a pure creative process. Building an optimal Web site requires that you test many variations of the site.

Here’s an introduction to some basic concepts of site optimization.

A/B Splits

A/B split testing is one of the most common forms of site optimization. In A/B tests, two variations of a site or page receive traffic divided equally between the two versions. A/B split testing requires that your Web application serve two unique versions of a page on a dynamic basis.

Once you have a split test set up, you run it for a period long enough to obtain a statistically significant amount of data. You can then look at the data and see what version performs better and use that version going forward. In some cases, software can automatically serve the winning landing page. To get the best results, A/B split tests should be done continuously in an evolutionary process.

Multivariate Testing

As you might predict, multivariate or multi-variable testing tests many different variables simultaneously. However, it’s impossible to compare all possible versions of a site or landing page at the same time.

Instead, test a few selected versions, and then based on that, try to predict which combination of the variables will deliver the best results. Implementing a multivariate test requires a substantial background in statistics. In addition, the number of variables you can successfully manage in this type of testing may still be somewhat limited.

Variations on Multivariate

A number of companies use proprietary technology to do sophisticated multivariate testing. One technology and professional services company, SiteTuners, has developed TuningEngine software that executes complex multivariate testing campaigns in real-time. It enables precise measurement of a large array of variables. SiteTuners has posted a case study featuring RealAge, the healthcare site just purchased on Monday by Hearst Magazines.

Optimization Case Study

RealAge.com is a site that obtains input about individuals and their lifestyles to calculate their “true” biological age, a number influenced by the wear, tear, and care of a person’s body over time.

The site has received acclaim, and has been featured by major news media including CNN, Fox News, The New York Times, and Good Morning America, among others. According to the case study, when RealAge got in touch with SiteTuners, it already believed it was doing a good job. RealAge had performed many tests and boasted a double-digit conversion rate.

Nonetheless, it wanted to see if the technology company could do better. SiteTuners helped RealAge identify 12 different variables that could potentially affect conversion. These variables resulted in 552,960 different possible versions of the site, far more than could be managed by any conventional means.

Once the testing was done, SiteTuners had identified a version with a 40 percent higher conversion rate than the original site, validated by the head -to-head test. This type of improvement would likely change the economics of any business. Moreover, RealAge had reportedly been optimizing its site already using more traditional means.


Ultimately, you can’t really know what’s in the head of visitors who come to your site. What’s most important, then? You know your products and services better than anyone. Visitors want to learn what you know, or at least a small portion of what you know.

Visitors to your site are also generally coming there to perform a specific task. They want to complete their task and move on. Many things influence their willingness and ability to complete that task on your site. It can be as simple as the call to action, background colors, title tags, tagline, site navigation, or any number of minor tweaks.

There really is a nearly endless list of possibilities that can have an impact on conversions and revenue. In addition, all variables interact with one another. To get the optimum results for your site, set up different scenarios and test them. The ROI on your investment will likely be significant.

Related reading

A primer to forecasting the value of SEO
How to get started with Data-Driven Attribution in Google Analytics
Conversation Mapping