MarketingExperiments Blog: Research-driven optimization, testing, and marketing ideas

We just received confirmation of a 133% increase in conversion for a recently completed test.

It’s one of hundreds of positive results we’ve seen by testing “calls to action” but only the second time we’ve seen it happen “on accident” while we were running a basket recovery test.

Here’s the setting. . . . We have a subscription based site with a free trial that we have been testing for months. There have been quite a few significant bumps up in conversion and we had yet to run a basket recovery test which has been a never-fail winner in terms of increasing conversion through recovering lost subscribers. The setup for the test was simple: just add an email capture field to the current call to action, give the customers a reason to enter their email address, and follow up with a friendly customer service email if they do not complete the sign-up process.

Typically, adding the email capture field has a negative impact on direct conversion (people signing up on their initial visit) but you make up for it plus some with the basket recovery emails that follow up if they leave the site before completing.

Here is a very typical set of results:

Conversion before basket recovery: 1.21%

Conversion after basket recovery (but before recovery emails sent): 1.09%

Conversion after basket recovery (after emails sent): 1.45%

Most companies would gladly take a 10% hit in direct conversion for a 20% increase in overall conversion. In this test, the results were drastically different and counter intuitive.

Instead of direct conversion decreasing a little, it more than doubled. It’s a rare case where adding friction and anxiety with the email capture field actually increases conversion.

Here are the results:

Uniques to offer page without email capture: 32,782

Free trial sign-ups without email capture: 222

Conversion: .68%

Uniques to offer page with email capture: 30,479

Free trial signups with email capture: 480

Conversion: 1.57%

Increase 133%

If you are scratching your head saying, “Add an email capture field next to my call to action and my conversion will go up,” it’s not quite that easy. We’ve had precisely this same result at least one time before, so let me spell out the similarities so you can determine if you should test this with your own site. An ideal site for this test (one that would most likely get a positive result):

• Is monthly subscription based

• Offers a free trial (or a solid money back guarantee)

• Uses longer copy-style pages where the call to action is typically at the end of the content

• Has a relatively low monthly cost (both are under $30 monthly)

• Is currently using a button or text for their call to action (not a form).

We did not think we would get this result from the actual changes in the call to action. It was totally unexpected. When you are running call-to-action tests with your own site, it’s essential that you test variations that are counter intuitive and just keep testing. We stumbled upon this one, but some of our greatest improvements have been because of precisely the same thing: We were testing for one thing and instead got a result from another that was unexpected.

P.S. If you run this test or have run a similar test in the past with little or no results, let me know and I may be able to take a look and make a few suggestions. Just make a comment.

Site Design

If you have any of the latest versions of SKYPE installed on your computer you may have noticed a little feature that converts any phone number on any website you may be browsing into a clickable phone number. The way it works is that it actually inserts CSS into the web page you are looking at to convert the text into a clickable SKYPE phone number.

Is this having an effect on your marketing?

We don’t really know, but one thing we do know… users that have this plugin will definitely notice ads that include phone numbers over ads that do not because of the extra emphasis given to the ad with the SKYPE code. Another interesting point is that when you click the number it does not go to the actual landing page on the systems we tested it with which leaves Google with lost click revenue and companies that will have a harder if not impossible time of tracking customers who call by clicking that link (unless they use a special phone number just for PPC ads). We are pointing out PPC but this goes for any site that displays phone numbers with html (it can’t read text on images).

Marketing Insights

By Jimmy Ellis
Director of Optimization Research,
Marketing Experiments

1. Build a new headline for a high traffic landing page.
2. Build a new ad in a high volume PPC campaign ad group.
3. Take one larger keyword group in a PPC campaign and split it into multiple, more specific ad groups
4. Sign up for Google Analytics (it’s free). If you already have it, take a look at your top exit points and figure out why customers are leaving on those pages. (http://analytics.google.com)
5. Add a new testimonial to a high volume landing page. If you don’t have any, send a personal email to some of your repeat customers and ask them about their purchasing experience or about the product/service they purchased.
6. Build a new button for your site (a call to action such as “buy now,” “buy,” “add to cart,” ). In many of our tests, “add to cart” out- performs “buy now” and similar buttons.
7. Add an incentive to subscribe to your email list (discount, special report, gift, etc). If you already have one, test a new incentive to try and improve your results.
8. Run a website speed test on your site (maybe your home page). You can do it free right here: (http://www.websiteoptimization.com/services/analyze/). Take a look at the results and try to reduce the size of your site by cleaning up your html, your css, or optimizing your images for optimal load times.
9. Drop or increase your prices. If you have never price tested, then you have never maximized the revenue of your current sales. Pick a high volume product or service, mark it up or down 10%, and watch those numbers. We prefer a/b split tests but if you don’t have a way to do it, a sequential test will usually give you enough information to tell if it’s having an impact on your revenue.
10. Place an order on your own site. When is the last time you did that? I can almost guarantee you won’t love it. Put yourself in your customer’s shoes and ask yourself, “How could that have been easier?“ Then have your designer test changes based on your input.

Practical Application