Boost Conversions Step 4: Run A/B Tests, Tweak, and Refine

This guest post is by the Web Marketing Ninja.

When it comes to conversion rate optimization, it’s easy to read about, and think about.

But when it comes to actually running a test, most people are at a loss.

It’s not that we don’t believe in testing; it’s that there’s barely enough time in the day to set up those key pages once, let alone set up variants, implement a test, measure, refine the pages, and test again. Trust me—I’ve been there!

But as we’re nearing the end of this series of posts about boosting conversions, I’m hoping you’re all fired up!

I’m going to use that motivation to push you to finally run that first test—a simple A/B test. In this post, I’ll run you , step by step, through a simple test that:

won’t cost you a cent
takes less than an hour of your time to set up
gives you that all-important glimpse of what testing can actually do for your blog.

I’ll bet once you’ve cracked that first A/B testing nut, you’ll become a testing junkie like me. And your conversion rates with never be the same—hopefully, they’ll be much better!

So let’s get testing.

1. Choosing a page

First things first—let’s pick a page to test.

In the second post in this series, Darren talked about reviewing your conversion funnel. That may have given you a few ideas about pages you could test—maybe they’re some of the pages you reworked after reading Tommy’s post yesterday.

My basic approach is, if you’ve got a sales or signup page that gets traffic, test that. (It’s likely to be on your list anyway.) If you don’t, pick your Contact page instead. Or, if you’re feeling brave you can go for the biggest bang for buck and test a “money page.”

2. Working out what to test

Our second step is to figure out what to test. When I’m looking at a page I want to test for the first time, I ask these six questions:

Can everyone access it? We’re talking here about accessibility.
Can everyone use it? Usability is the key for complex processes.
Does it work? It should—on all browsers, mobile devices, non-javascript browsers, and so on. Don’t forget to consider page load speeds as well.
How does it look? Does is communicate the mood you want it to?
How well does is tell the story? Do the words engage users and drive the actions you want?

Ask these questions about any web page. and you’ll end up with a long list of stuff you can test, but for now, let’s start with a headline—a big part of telling the story, and probably a fairly strong element in any sales or signup conversion. It’s also something that Tommy was eager to test yesterday, in his third conversion goal, which was to get more high-quality leads.

As this is an A/B test, you need to come up with just one alternative to the page’s original headline. If one email can have over 500 different subject lines then I think we can probably come up with one.

Now we’ve got a page, we’ve got our original headline, and we’ve got an alternative headline. Let’s start our test!

3. Setting up the test

You can use a few different applications to run web page tests—some free, some not. To keep things simple, we’ll use Google Website Optimiser—one of the free options.

In order to use this tool, we first need to set up a couple of things.

We need a publicly viewable version of your original page, and the one you want to test with the new headline. And you’ll need them at two separate URLS—it might be problogger.net/salespage.php and problogger.net/salespage1.php. These URLS will depend on the CMS or blog technology you’re using and your site structure, of course.
We need access to a page that appears aftera user completes your goal action. So, in the case of a contact form, this page would be your “thanks, your message has been received” conformation page.If you’re testing a sales page, this can be a little more tricky. Ideally you’d have access to the page that confirms that the user’s purchase has been successful. If you can’t access that page, you might have to settle for the page that appears when someone clicks on of your Buy Now links.(Note that there are ways around this problem, however you might need some technical assistance to access them. In this case, I would recommend you look at a service like Optimizely/, but it’s not cheap. The upside is that once you set it up, creating tests is extremely easy.)

Once you’ve got all of that done, sign up to Website Optimiser. Once you’ve signed up you should see a page like the one below. Click the link to start your experiment.

You’ll then be asked what type of test you want to run. Pick the A/B Test.

You’ll then be asked to get your test pages and your conversion page ready. We’ve already done that, so we can confirm and move to the next step.

Next, you’ll need to enter a name.

Include the links to the original page, and the version you want to test.

Finally, paste in the link to your goal or conversion page.

Once you’ve completed all the fields, click Continue.

The next step is the most technical. You need to put a special piece of code into your original page, your test page, and your conversion page. (You can read more about the code snippets themselves here.)

If you’re using WordPress, there’s a handy plugin that will allow you to do this pretty easily, called Google Website Optimizer for WordPress.

Once it’s activated you’ll see a spot under each page and post to enable testing—add your special code in there. If you’re confident with editing the tags on particular pages, great. If you’re not using WordPress, you’re not technically minded, and you can’t find a Website Optimizer plugin for your CMS, you might need to ask nicely for some help.

I’m going to move on, assuming that you’ve got the codes in place. Next, you’ll need to validate them:

If the validation’s all good, you’ll get a screen that looks like this:

Click OK, then click Next. You’ll arrive at the final conformation screen, where you can preview or start your experiment.

Once you hit Start, you can sit back and relax for a bit: you’re now testing! After a few hours some of your preliminary results will start to come through. When you log into Website Optimizer you should see your experiment listed. To see the results, click on the View Report link. The report shows you how the two pages are performing against each other.

4. Deciding the winner

You can expect to see some wild fluctuations in the data initially, so it’s important not to decide on a winner to quickly—let the data smooth out over time. In the case shown above, the results came in pretty even—and this is a test I ran over four months!

Most testing platforms will have an algorithm to let you know how confident they are that one version is beating another. In the case of Website Optimizer, it’s called a “high-confidence winner.” In the case of slight changes, it can take a while for a call to be made. You can either wait, or pick your own moment and move on. It’s really up to you.

Personally, I’ve made calls on tests that have only run for three days, and waited for some that have run over months and months. As your experience in testing grows, so will your confidence in making calls.

What to expect from your test

Within your tests, you’ll probably experience one of three things:

Your new headline wins.
Your original headline wins.
The result is too close to tell.

In the first case, you’ve hopefully got a great understanding of the progress you can make with testing.

If your original headline wins, you’ve actually also made a small step forward: you’ve proven that your current headline is better than at least one other option—but I’m sure there’s a bunch more to try!

If it’s a to close to tell results, then, as is the case if the original wins, it’s time to think up some new headlines.

So hopefully you’re all able to identify, set up, run, and report on a simple A/B test. Even better, I hope you’ve found it so easy that you’re ready and raring to start your next test. Because if you’re happy with good, then produce. But if you aspire to great, then produce, test, iterate, test again—and you just might get there.

And that’s the key point here: to continuously improve your blog’s conversion rates for paid or unpaid offers, you really need to have in place an ongoing system of refinement that’s based on trial and testing.

Once you’ve got a handle on that, you’ll be able to go back and apply the four steps for boosting conversions—reviewing your offer, revisiting your conversion funnel, revamping your communications, and running A/B tests—more broadly, to every segment of your audience. That’s what we’ll be looking at later today, in the final part of this series. Don’t miss it!

Stay tuned for more posts by the Web Marketing Ninja—author of The Blogger’s Guide to Online Marketing, and a professional online marketer for a major web brand. Follow the Web Marketing Ninja on Twitter.