If you’re a savvy marketer, you’re probably testing your web pages. You need to know when visitors click on your page links, that you’re getting optins, making sales, or whatever else your page is intended to do, and you should want to see those elements improve over time. If you’re new to marketing online or if you’re a lazy marketer, you’re probably doing no testing at all.
Shame on you!
This is a HUGE mistake. It’s important to run tests continuously to improve your performance and to meet your goals as quickly as possible.
There are two types of testing. First, you have the simple A/B split test, where one page is pitted against another. Only ONE element of the page should be tested, however, or you won’t know what’s working and what’s not. So, you can test your headline, your sub-head, your deck copy, your optin form, images, etc. but only one of those elements at a time.
Let’s say you want to test your headline. You add another page to your site that is identical in every way to your “control” or original page, except it has a different headline. Then, you sit back and watch what happens.
If the new page wins out over the control, then the new page becomes the control page and you start your test over, using a third headline. You can do this infinitely or until you reach your goal.
There is a way to make this easy for you. Write 50 headlines for your page at the same time and then, it will be easy for you to test. Sometimes we’re so busy that the thought of writing ONE headline is daunting, but if you have them ready to go… it’s not such a big deal.
Multivariate testing involves using software that will allow you to test more than one element on your page at a time. This is the quickest way to test, but it can be confusing, especially if you’re using Google Analytics to do the testing for you.
With Website Optimizer, which went away last year, it was fairly straightforward. One page, add code, and BOOM! Done.
But with Analytics “Content Experiments,” it’s not quite as simple or as fast to set up. Instead of adding special coding to one page, now you have to set up multiple pages to get the same results.
You can, however, test three or more headlines at a time. It really depends on how much traffic you’re getting. If it’s not a lot, stick to two versions of the page so you can find a winner more quickly.
But if you want to test three headlines, you need to set up three pages, each with a different headline and the proper experiment code has to be added to each page. And let’s say you want to test your optin box at the same time. You need to set up another test with two or more pages to run that test at the same time.
Getting this all set up is not really that hard to do, but it definitely takes more time. You also have to set goals in Analytics to make the test work. Of course, goals are good, and you should have them set up anyway; however, you may have to create different goals for multivariate tests.
Plus, it is über important that you do these experiments properly or it will affect your SEO.
Since you’re only changing one small part of your page, the two pages that aren’t your control or original page are duplicate content on your own website. This can be a disaster! You know that Google is anti-dupe content now, but when it’s on your own website, you’ll see your rankings fall like a rock tossed off the Empire State Building.
Can you say, “Gone in 60 seconds?” I knew you could.
Six years ago, I went to bed, happy in knowing that a site I was working on was at #3 in the SERPs at Google for a very competitive term. When I woke up and checked on positions the following day, I saw that the page had fallen to position 563! (That number is burned into my brain.) Of course, I freaked out. The first thing that goes through your mind is that you’ve been Google smacked.
But we weren’t. As I’m looking around the site trying to figure this all out, I realized that the site manager had taken content from the blog, which lived on the same domain, and posted it to the main site, as well. Bingo! The day or two after that duplicate content was removed, our #3 ranking returned. Whew!
That was then; this is now.
In a post-Panda world, duplicate content anywhere on the Web is a not-so-good thing.
So, what to do?
You’ll need to use the “rel=canonical” tag in the header information on each of the test pages, pointing back to your control page. This is the best way to handle the situation, and what Google recommends. Using “noindex” on the test pages can cause problems. Google says in their Webmaster Tools article “Website testing & Google search”:
“Using noindex rather than rel=“canonical” in such a situation can sometimes have unexpected effects (e.g., if for some reason we choose one of the variant URLs as the canonical, the ‘original’ URL might also get dropped from the index since it would get treated as a duplicate).”
You don’t want that to happen, so adding a simple line of code to your test pages is the best answer.
Whatever you do, don’t try to use cloaking. Google really gets its panties in a twist about that! Cloaking is when you send spiders to one page and visitors to another. Never do that or you could find yourself de-indexed.
If you’re running a redirect on your control page, where software or coding sends visitors to either control, test A or test B, then you may want to use a redirect from the control pages to the test pages. If you do that, be sure to name it a “302” or temporary redirect, rather than a “301” redirect, which is permanent. The danger here is that your control page could be dropped from the index if you tell spiders that the redirect is permanent. Again, you don’t want that to happen! What if it really IS the best page?
The last thing you don’t want to do is to run the test too long. Rule of thumb is 1,000 unique visitors, but marketers vary testing times according to the results they experience.
One way to see if you really have data that is meaningful is to use a statistical significance calculator. There are several available online, but many of them have to do with random samplings tests, rather than A/B tests.
One calculator that is set up specifically for A/B testing is KissMetric’s Data Driven A/B Significance Test which you can see here. (You can also get some pretty sweet conversion and other tools there.)
The Bottom Line
The bottom line is that you MUST be testing. If you aren’t, you could be losing a lot of business. But if you are testing, whether it’s multivariate or A/B split, keep SEO in mind. Without the proper information for spiders in place, you could be harming your site in search and not even know it.
Don’t do that!