Best bets for site tests

Apr 01, 2008 9:30 PM  By

Nearly every Website team tracks site conversion. But few really do anything about it. That is, few use scientific testing to discover real improvements. Surprisingly, this is true even among organizations with decades of direct marketing experience. Too often, companies change their sites based solely on creative “instinct,” IT availability, executive fiat, or what a few people think the numbers might be saying.

Then, following the roll out of a new design — often long in coming — the “before and after” measurements are taken. If the numbers look good, victory is declared. If not, it’s back to the drawing board.

But what are the chances of building a more effective site — even optimizing a single landing page — if you don’t know which elements boost conversion and which elements depress it? And how much time can you afford to spend guessing?

Site testing offers an alternative by showing you what works, precisely and quickly — with real customers.

Free MVT with Google’s Website Optimizer

Several premium testing packages are available today, but recently Google has shifted the testing landscape with the launch of its powerful Website Optimizer tool, a testing platform available free of charge to Google advertisers (in other words, free to virtually any multichannel retailer).

This tool changes the testing game much as the release of Google Analytics changed the worlds of those providing and consuming site analytics software.

With the cost of testing technology dramatically reduced, site owners can focus solely on where and what to test, optimizing their sites to sell.

Google’s Website Optimizer provides online marketers with an advanced statistical testing method: multivariate testing (MVT). Think of MVT as traditional A/B split testing on steroids. MVT lets you run statistically significant tests of multiple site elements in multiple combinations — often quite quickly. To run comparable tests with traditional A/B testing would require more site traffic, more time, and more creative effort.

Multivariate testing lets you pinpoint exactly which copy, graphics, and layouts drive more sales. This type of testing is changing the way successful Websites are created: Bloated, slow and risky wholesale site redesigns are out; nimble, incremental, and provable site changes are in.

While Google’s free tool naturally lacks some of the functionality of premium testing packages, for most online retailers, it’s an option to be seriously evaluated and a great vehicle for getting started. Google is committed to continuously improving the tool and actively solicits feedback to this end. And because the tool has been embraced by both agencies and the developer community, several effective third-party “plug-ins” help enhance the tool’s usefulness.

Ultimately, successful tests are based on insights and sweat equity, not software.

Regardless of the platform a retailer chooses, it’s getting riskier to sit on the sidelines. A low-cost tool means your competition will soon be testing, even if you’re not. This has implications that go beyond your site’s ability to close the sale — it has direct impact on the profitability of your customer acquisition programs, too.

Consider your paid search marketing. A fine-tuned PPC program drives qualified traffic to your site, cost efficiently. But competition continues to increase, and industry cost per click keeps rising. Your ability to wring value from your site visits is critical. Increased sales per visit would let you bid more, within your targeted efficiency. Higher bids would generate more traffic. More traffic and higher conversion would drive more sales and more profits. At day’s end, the retailer with the most effective site will win the PPC wars.

What to test

Success in testing is less about finding a single silver bullet and more about finding the right combinations of the right elements. Tools like Google’s makes it easy for your site to serve and track these combinations on the fly, but you still need insight to pick the ingredients and write the recipe.

Here are just a few of the basic elements you’ll be testing:

  • Headlines: Among the fastest and most effective elements to test.
  • Images: Test people vs. product, solo product vs. lifestyle shots.
  • Copy: Test copy length, test calls to action, test link text, test voice.
  • Product presentation: Test product density/page, test single heroes vs. multiple favorites.
  • Layout: Test to find the optimal grids for key pages such as home page, category and site search results. Test editorial content presented in a single long page vs. a series of short pages.
  • Global elements: Test size, color, shape and wording of global buttons like “add to cart.” Test product organization and navigation schemes. As you mix your ingredients, beware of “bland recipes” — they often produce negligible results. When my firm works with clients, we recommend “testing shouts, not whispers.” Solicit new perspectives and push yourself to test “out of the box” ideas.

Finding high impact pages to test

What to test is just one piece of the puzzle; where you test also matters. Even with a free tool, your own time and resources must be spent wisely. Apply your effort to pages where changes can have real impact on your business.

Looking for a strong place to start? Use your paid search reports to identify underperforming pages. Look for your high-cost terms, and pages with high traffic and low conversion. As you broaden your scope, your testing candidates will include top landing and entry pages, site search results pages, cart, and checkout. Mining your site analytics and listening for customer pain points can help you identify the best opportunities for your own site.

Among the Web metrics I find most useful is the bounce rate, the percentage of users who arrive at your site and then leave within seconds. Take a look at the high-traffic, high bounce rate pages on your site. Why do visitors show up and promptly leave? Compare the content of these pages to best practices, to your competitors, and to more successful pages on your own site, and start testing.

Create a testing culture in your company

In many organizations, the most significant challenges to testing aren’t technical, or a dearth of good ideas. Instead, they’re cultural. Many companies simply aren’t comfortable making data-driven Website changes, preferring to continue to lean heavily on internal opinions, often voiced top-down.

For your company to reap the economic benefits of site testing, you need to create an environment where it’s okay for anybody to be right and anybody to be wrong.

Here are some tips on how you can turn your team into a band of energized and profitable site testing ninjas:

Test continuously Accept the stark reality that many, many tests produce null results. But don’t let that stop you: Let it force you to build momentum. The only path to success is with continuous experimentation. Create a site testing calendar to drive your testing regimen. The more you test, the greater your chances of moving the needle.

Prepare to be humbled Testing often proves the smartest people and best ideas wrong. That’s a good thing: Your CEO, copywriter, call center folks, analysts, designers, programmers and merchants all have a chance to come up with a winning idea or a flop.

Appoint a site testing manager with solid analytic (and evangelical) skills and task them with building your cross-functional testing team. (These people keep their day jobs, but carve off time for testing, too.)

Follow a testing roadmap. Each test should begin with a clearly stated hypothesis identifying what will be tested and the expected business outcome.

Be clear about the pages and content to be tested, the form of the test and metrics for your anticipated success. Of course, your roadmap should also include a step for quickly rolling out successful changes after your test is finished.

Document your tests. Create brief write-ups of what was tested, why, and the results that followed. This makes it easier to share learnings across teams, and across channels. For example, the top-three combinations of headline, image and copy from the recent homepage test could save your print team time and increase chances of success when used to narrow the field for your next catalog cover test.

Look beyond your four walls for good ideas. Talk to other retailers, and also consider tapping an agency as you launch your testing process. In addition to fresh perspective, a good agency can bring design resources to supplement your internal team, custom IT to help you get more from your test platform, and a repertoire of testing ideas that they’ve seen produce results across a variety of sites.

To start implementing winning results, you need to start testing now. Your ability to test can provide a true competitive advantage with measurable impact on your top and bottom lines.


Larry Becker is vice president and principal, Website effectiveness at the Rimm-Kaufman Group (www.rimmkaufman.com), an online marketing agency offering Website consulting and paid search services.