March 2010 – Column: Best Practices in Online Marketing

How Can One Website Outperform Another?

Our clients tell us that every web marketing company out there claims their DRTV landing page websites have the best performance. By “best performance,” I think we all understand it as a website that will generate the most revenue per visitor or, put another way, the highest conversion rate and the highest average sale.

I guess we (web marketers) all better say that if we want to get clients and stay in business. So I make the same pitch. But inevitably the client or prospect says, “You guys all say you have the best performance. So what makes yours better than the next guy’s or the one my friend built me. How can any one website outperform another?”

This is a good question. How can one website possibly outperform another? To the uninitiated, it may not even seem logical that this could be possible.

A MULTITUDE OF LITTLE THINGS
The truth is, we have done thousands of tests over many thousands of orders. In doing so, we have discovered a multitude of (seemingly) little things that affect performance and, as a consequence, we have discovered that one website truly can outperform another.

We have won and lost thousands of tests against ourselves over the past three to four years, and we continue to discover new things that both help and hinder performance. And knowledge of each is equally important.

Speaking of this “multitude of little things,” can you guess what a few of them are? Can you estimate how significant of an impact they can have on your business? Without getting so specific as to give away trade secrets, we know that by doing one certain thing on a landing page, we can improve your conversion rate by 50 percent or more.

Did you know that by structuring the presentation of your offer in a certain way, you can improve your RPO significantly?

Did you know that placing text in a certain location or picking the wrong headline can cut your response rates in half?

Two websites rarely perform exactly the same even if you change only small, seemingly innocuous elements. In fact, I have seen the scenario with two otherwise identical websites, where two–two!–words were added to a page on one of them–and in 10-point type–and response dropped by 80 percent!

So believe me, seemingly similar websites do perform quite differently, and the difference in results can have a dramatic impact on your bottom line. In this sort of environment, there is but one truism: you gotta test, and test constantly!

Think about it, would you have the guy down the street who has a video camera shoot your commercial instead of hiring a talented producer with the experience needed to make it a success? Of course not!

Why? Because you know that two different commercials for the same product can yield drastically different results. The one produced by your neighbor and his camcorder would never get the same response as the professionally shot and edited version from a pro.

The same is true on the web.

If you are only running one version of your website–in other words, if you’re not testing–you’re getting out-hustled and out-performed by the guy down the street. Assuming you have multiple website creatives for the same product running on a single platform, here are the four fundamentals you need to have in place to determine which one is the best-performing site.

You need a traffic splitter and an accurate reporting mechanism.

You then need some kind of testing plan to determine which items to test. (This would include things like the positioning of different site elements, headlines, form configurations, Flash vs. HTML, plus dozens of other items.)

You need a platform on which to conduct progressive A/B split or multivariate tests against a control, continually moving forward with the winners of each test.

You also need a technical platform specifically designed to execute and report usable test data. We have one we call AutoCart.

Once you have these basic elements: a traffic splitter, parties focused on accurate results, a testing plan and a testing platform, you are ready to begin.

As an example, let’s say you want to begin by testing something simple like background coloration or the positioning of the video on your website. To test the video player, you would build four or five versions of your site with the video player in various locations on the page. Then, run a minimum of 1,000 orders through the splitter and you would notice that at least one of the five sites will provide higher conversion and revenue. (In a recent test case we conducted, the location of the video that provided this boost was the lower left position.) Next, you would apply this to another five websites for other products and see if the results replicated. If they did, you would know you were onto something and could expand your testing process to additional items. Once you tested those, you could move on to the next ones. The process needs to be continual.

Once you have done this process for three to four years, conducted over 1,000 tests on all elements of website designs, layouts, buttons, logos, seals, copy, upsell configurations, positioning, etc., we think you’ll have a different view on whether all websites perform the same.

Bob Greenstone is CEO of Permission Interactive Inc. in San Diego. He can be reached at (619) 708-7456 or at bob@permissioninteractive.com.