This is the second part of a post that appeared earlier this week.
There is no better way to increase conversion rates than to carry out website testing. Whether you opt for multivariate or A/B testing protocols, serving different versions of your site to measure response means you can tailor each page to best meet user expectations and demand. But while it testing is important for conversions and performance, it must be carried out in line with Google’s guidelines to guarantee that your efforts to better the site do not come at the cost of your relationship with the search engine. So, how can you be sure that your testing stays on the right side of the rule book? Here are three “must-do” items:
1. Avoid Cloaking
Cloaking refers to the practice of showing Googlebot one version of the page and the user a second. Cloaking is a black hat SEO technique and high on Google’s list of dislikes. By showing Google one version and your visitor a second, the search engine is very likely to think you’re attempting to fool it as part of a misguided attempt to gain better keyword rankings. Embarking on a testing schedule without keeping this in mind means you are more likely to fall foul of the cloaking police, even if your website testing has nothing at all to do with black hat optimization.
To avoid appearing to be cloaking, you’ll need to make sure that the version of the page served to the user is not based on user-agent. This means that the user-agent Googlebot should be treated just like any other visitor and be served both versions of the page; the original and the test version.
2. Set a Time Limit
Although it is fine to test any page on your site and to treat testing as part of your ongoing commitment to improving your site conversions, it’s important to put a time limit on each test. Google advises that the time required to conduct a useful test will vary depending on a number of factors, not least of which is the conversion rate of the page, however the test should be halted when you have enough data to draw reasonable conclusions. If you opt for software-aided testing of your site pages, a good package will automatically tell you when you have enough data to process. At this point, the testing can be stopped and the most successful version of the page set live.
When the data gathered is sufficient, the elements which make up the test such as alternative URLs when A/B testing, testing script and page mark up should be removed and the page returned to its non-testing state. Google warns that if they see the test parameters on a site for what they consider to be an unnecessarily long time, it will consider the whole exercise an attempt to manipulate rankings and take further action as necessary. As this further action could include de-indexing or keyword penalties, it’s important to be on the ball and pause tests as soon as possible.
3. Choose Redirects Carefully
The A/B method of testing a page requires two variations of the URL in order to split website traffic between the pair of test pages. When this method is used, redirects become necessary. When the visitor needs to be directed from the original URL to the test URL, a 302 redirect should be used. This tells Google the change is temporary and will only be used for as long as the test is running. Using a 302 rather than a 301 is important because the 302 tells the search engines not to replace the original, indexed page with the temporary page which may only be published for a matter of days. Using a 301 instead of a 302 can impact search positions and may cause a loss of visibility, particularly when the temporary test page is removed.
The rel=”canonical” link attribute can also placed on each alternate URL created for testing purposes to advise the search engine that the original page is the preferred version. Google recommends rel=”canonical” be used ahead of a no index meta tag on the test page in this situation.
Image: Testing Disasters by Shutterstock