Learn how to test for conversion rate optimization with A/B and MVT testing.
- Every marketer knows that there's always room to improve when it comes to the effectiveness of our programs, and that's especially true in the realm of digital. So much so in fact that there's an entire practice dedicated to optimizing how people accomplish what we want them to. It's called conversion rate optimization, or CRO for short. One big part of this is structured testing, where we create different versions of what we're putting in front of our audience, and we measure how each of those versions performs. Now this isn't a new thing by any means.
Traditional marketers were sending out different versions of direct mail pieces and catalogs decades ago, and they were tracking which ones had the better response rates. The advent of digital just means that we can run these kinds of tests on more of our marketing assets more efficiently and on a much bigger scale. Among the most popular forms of testing used by digital marketers these days is website and landing page testing. There are lots of tools out there that can help you do this, and once again, they all offer different features at different price points.
You've got tools like Google's Content Experiments that's embedded right into Google Analytics. Visual Website Optimizer, Unbounce, Cubit, AB Tasty and many, many more, and towards the enterprise side of the house, you'll find tools like Optimizely, Maxymiser, Adobe's Target, HP's Optimost, SiteSpect, HiConversion, Monetate, and others, but regardless of which tool you find is right for you, generally they all allow you to plan, create, launch, and measure your experiments. There are two general types that are supported by just about every tool out there, A/B and multivariate tests.
An A/B test is typically about testing completely different versions of things. Think about a landing page for a digital campaign that's built specifically to catch someone that's clicking on an ad. You might want to try out all kinds of different things to see what works the best. With this kind of testing, you can pit version A against version B, or of course, you could have version C and D, and as many more as you want. The tool will decide how to split people up so that different users get directed to different versions, and it will record what happens.
Now after the test has run for a while, you can see the reports where all the statistics and the numbers have been crunched, and you'll see your winners and your losers. The multivariate experiment is a little different. This is when you're testing different components within a page or website to see not only how they impact conversion rates, but also how the different components interact with each other. Think about a landing page that has an image, a headline, and finally a button that you want people to click. With a multivariate test, you could try out a couple different images, some different headlines, and different button styles, and by doing this, you'll find the best combination of all of these elements with respect to conversion.
Once again, the tool will help you load up the different variations, it will split up the traffic for you, and it will track and crunch all the numbers to show you how things perform. There's a lot more to this of course, but again, let's focus on how this can integrate with the rest of the marketing technology stack. First, integrating a testing platform with a web analytics platform can give you another level of depth in your analysis. Rather than just focusing on the big, general groups of people who saw one variation or another, you can use all the web analytics data to examine different user segments, and different user types.
Maybe that headline that looks so good in the test results was only good for people who had never been to the site before. Or, maybe a certain image works really well with people from a certain geography, or that come as a result of seeing a specific ad variation. Now you can do a lot of this kind of analysis right in a web analytics tool, but to dive deeper, you'll likely be feeding all this integrated data into a statistical analysis tool, using something like a SAS or SPSS, Minitab, JMP, R, or any of a number of others.
Last, websites aren't the only thing that can be tested. Just about any channel where you have the ability to control what your audience is seeing is a good candidate. Today, unless you have a conversion and click rate of 100% with infinite levels of profit, you should never be sending out just one version of an e-mail, or loading up just one version of a text or a display ad. Testing is a great path towards that virtuous cycle of continuous improvement, and understanding the tools and the data inputs and outputs that allow you to run, analyze, and learn from your tests is a clear advantage to any marketing organization.
NOTE: While specific software and platforms aren't endorsed, you will see how tools like a customer relationship management system and web analytics work in a successful marketing mix.
- What is digital marketing?
- Understanding the marketing data being generated
- Reaching customers via digital channels like social, search, and display
- Working with digital experiences
- Selling online with ecommerce
- Going mobile
- Measuring and optimizing with testing and analytics
- Running and operating a business with technology
- Storing and extracting data
- Learning and predicting with data exploration and modeling