Companies waste an estimated $6.6 billion on unused software in the U.S. every year. With more options than ever, finding the right software for your dealership can be a challenge. How can you cut through the clutter and make sure your software dollars are well-spent? Download your free step-by-step guide to successfully navigating the software jungle. DOWNLOAD GUIDE
There are three rules when it comes to properly A/B testing the various components of your website and the tools that drive it. We use them and we know there are a handful of other vendors who use them as well, but you don't have to be a vendor to take advantage of them. Dealers can do this type of testing as well.
This should not be confused with multi-variant testing, a technique often used in PPC and landing page optimiztion that gives multiple variations of a page to see which converts best. With component A/B testing, you're looking for micro-improvements rather than finalist selection.
Here are the rules for basic A/B testing. There are others, but if you make sure to follow these, you'll have the best chance of success:
One of the most common mistakes made to both widgets and entire web pages is the concept that "it doesn't work, so let's change it." With that perspective, it's appealing to make several changes. This is the wrong approach.
You can learn a lot more about pages and widgets by testing one change at a time. If you know that a page is not converting and your hypothesis is that there are potentially three things wrong with the page, don't change all three. If you do and the results improve, you won't know which change made the most difference. Even worse, one of the changes could be very positive but one or two of the others could be negative, so you might have a winning change you could apply elsewhere that you ignore because the other two changes confused the numbers.
It's easy to get impatient. Proper A/B tests take time. You have to make sure you have a complete set of data before making any decisions.
Unfortunately, it's challenging to know when you have enough data. In our own data analysis model, we collect data for years. We make some changes after a short period of time, perhaps a couple of months worth of data, while other changes could take a longer period to get a good data set.
If you're testing widget placement, for example, you might see immediate results improve just by moving it from the bottom to the top of the page. Don't jump just because the data looks good anecdotally. Let it work its way through and continue to test for at least a month or two, depending on the volume of traffic that gets to see the widget.
This is the key to long-term success. You should have tests running on most components of your digital marketing at all times. In our world, a little plus a little plus a little can equal a lot. We always test because that's what it takes to improve our product. You should do the same.
It may seem tedious, but the results can be tremendous. For example, we've seen dealers test their pricing model to see whether round number prices ($24,888) work better than odd prices ($24, 827). You can test different variations of wording on contact forms. You can test whether it's better to say "hi" or "hello" in chat (yes, that's something that we've done and the results were pretty awesome).
The image above is a sampling of our chat operator images. We've seen a clear difference in lead conversion based strictly on the faces used in the chat console. Can you guess which of these was the best (by 7.94%) and which was the worst?
Take a guess and comment below.