In our last couple of posts on the topic of paid search we talked about click-through-rate and relevancy in your paid search campaigns. Both are very important when looking to get qualified traffic to your website and starting to get conversions.
Following those best practices will help you get good results. But, how do you know if those GOOD results are the BEST results? Chances are they aren’t. That’s where landing page testing comes into play.
In this blog post we’re going to discuss using A/B and multi-variant testing to maximize results from your paid search campaigns.
What is A/B Testing?
A/B testing is the act of taking a baseline version of a piece of creative (for the purpose of this blog, a landing page) and comparing it to a single-variable test version. For example, we might have a landing page with a big green “buy now” button and want to see if the response rate on the page is better if that button is red. We’ll then take a percentage of our landing page’s traffic and show a version of the page with the red button, and show an equal percentage our original green version.
We’ll measure our response rate on both and bingo, we’ll know which one performs better.
What is Multi-Variate Testing?
In simple terms, multi-variate testing is the same as A/B testing, except with many different items being tested on one page at the same time. A/B testing allows for only one variable to be tested at a time. For the beginning tester, there are more than enough possibilities and gains to be had using only A/B testing, so we’ll focus on multi-variate testing at a later date.
How can I test?
There are several applications available for landing page testing, ranging in price from free, to thousands of dollars monthly. Since free is always good, we’re big fans of Google’s Website Optimizer tool. Another good tool for both testing & creating landing pages is Unbounce. Unbounce comes with a small monthly fee, but allows you to easily create, manage & test landing page combinations from the same interface.
For someone currently doing no real testing, either of these solutions will provide you with an acceptable interface to start realizing increases in conversion rates with a/b testing.
What should I test?
Many people are surprised by the things on a website that might make a difference in conversion. Try testing things like:
- Page headline
- Call to Action
- Button location or size
- Button Colour
- Product Images/Images on Page
- Image size
- Location of Images
- Location of descriptive text
- Different versions of descriptive text
- If you’re running a form on the page, try testing less/more form elements and different submit button designs & calls to action
When it comes to testing, anything goes. Just remember that with a true A/B test, you only want to be testing one variable at a time. Op Ed recommends reading “Which Test Won” to get some great testing ideas & insight into results from testing. It’s also a great place to see just how good your marketing intuition is.
How often should I be testing?
Always be testing. Testing is not a one time project, it’s an ongoing battle and should always be part of your marketing plan. After you run one successful test, it might be tempting to say “great! We’ve increased conversions by 10%” and call it a day. Ask yourself again, “is good the best it can be?”. Try testing something else, and you might be rewarded with another incremental gain.
When is a test finished?
It’s difficult for us to give you an exact answer without knowing the details of your business (hey, why not contact us?). Just remember that you need to make sure the results of your test are statistically significant. It is never a good idea to make decisions based on limited data. Google Website Optimizer will help you with making sure the test has received enough data before determining a level of confidence in the results, but it is important to understand your results as they pertain to your business. In testing, just as in baseball, small sample sizes can lead to some poor decision making. (Eric Hinske’s contract with the Jays anyone?). Your test should receive sufficient traffic in relation to your site norms to be able to determine a clear winner.