Why A/B Testing is Like Eating an Elephant

I’m something of a NASCAR fanatic (I’ve got it pretty bad, actually) and one thing I’ve learned from the car setup guys is that when you test a race car, you change only one thing at a time. That way you know what works and what doesn’t.It’s pretty much the same with e-commerce. I was reminded of that when the folks at Future Now’s grokdotcom.com sent me their conversion rate newsletter with an article on “The ABCs of A/B Testing.”

This is useful stuff for the small e-commerce operator. That’s because all of us are trying to maximize sales and I know from experience that the temptation is to change a bunch of things all at once to bolster the bottom line whenever you encounter a little slump.

You want to try design changes, a new e-mail campaign, changes in your advertising, etc.

With race cars, you change something, then run a few laps, and see if you go faster.

With e-commerce, A/B testing — sometimes called an A/B split — is the simplest and easiest method of testing elements in your e-mails or on your Web site, according to grokdotcom.

“You divide your audience into two groups,” grokdotcom says. “You expose one group to the original version of whatever you are testing. You expose the other group to an alternative version, in which only one element has been changed. Then you track the results.”

It’s that simple.

FutureNow and grokdotcom are focused on turning visitors to a Web site into buyers or subscribers or sales leads and they talk a lot about conversion rates — converting those visitors into customers.

“Conversion is terribly misunderstood,” Future Now CEO Jeffrey Eisenberg told me. “Most folks forget that everything is made up of tiny little variables. People get too comfortable if they are profitable at a 2.2 percent conversion rate and break-even is 1.4 percent – they can’t imagine what’s happening to the 97.8 percent of the rest of the traffic. People get complacent …”

For e-commerce operators, Eisenberg recommends focusing efforts “earlier than the shopping cart pages, because you’ve already lost most of your customers who never got there.”

“The point isn’t to look at the final conversion rate, but at conversions from the micro-perspective,” Eisenberg said. “Each ‘click’ (by a visitor at your site) is a micro-conversion on the way to a macro-conversion.”

“I’d like to remind everyone of the old riddle: How do you eat an elephant? Bite by bite,” Eisenberg said.

“Every time your visitor is persuaded to delve one step deeper into their individual buying process, as reflected in a Web site architected for persuasion, they are completing a micro-conversion. Macro-conversions are obvious (sales, leads, subscriptions, etc.) and therefore easy to measure. However, if you neglect the micro-actions you are left with an elephant that seems too enormous to digest.”

The recent A/B article, written by Lisa Davis at Future Now, offers suggestions on implementing a testing process for changes on your site, such as:

  • Always be clear about what you are testing and how you are going to measure and interpret the results before you begin. You cannot measure success unless you know exactly what you are measuring.
  • Your test groups should be of similar size.
  • After the first test, you should always test against a control. The first time out, you are testing two unknowns, and you won’t be able to determine the better option until the results are in. Once they are in, then you have a benchmark, or control, against which you can measure subsequent changes.
  • Remember that you can accurately test only one element at a time. Even if they all seem necessary, changes need to be made individually so you can track effectively the result of the change.
  • If you are testing e-mails, send your test e-mails simultaneously to eliminate the timing variable.

The article says that maximizing your conversion rate is not simply a matter of making changes, it’s about making:

  1. the right changes,
  2. at the right time,
  3. in the right sequence, and then
  4. evaluating the results before continuing the process.

I think most NASCAR crew chiefs would agree with that.

“A/B testing is far from rocket science (and there are other more complicated and robust ways to test), but it has a sweet advantage: it isn’t complicated,” Davis writes. More importantly, it means you won’t have to make potentially expensive decisions based on your gut reaction. You’ll know, because you can say, “Here, look at the numbers!”

Adapted from ECommerce-Guide.com.

Must Read

Get the Free Newsletter!

Subscribe to Daily Tech Insider for top news, trends, and analysis.