Multivariate and A/B Testing without Hassle

ab testing
cross channel insights
Source: Google Analytics

One of the recent queries of the webmasters is whether or not the website tests, through multivariate and A/B, affects the site’s search performance or its ranking. This is one critical issue that had probably persuaded the Google analysts to act on the matter outrightly as what they did with the previous technical inconveniences.

Notably, analyst Susan Moskwa, through the Google’s webmaster central blog, expressed her team’s gratitude for the initiative of site experts in bringing up such concern immediately. As she noted it, the said testing is one efficient means to make certainty in what site administrators are introducing really have impacts to their users.

Nonetheless, in understanding the implications of the said activity, it is first to be noted that there are essential distinctions among the general and specific terms in testing. Primarily, as to the website testing, this is when a webmaster tries a lot of versions of his website and gather information on users way of responding to each of them, or in simple term, utilizing software in tracking which type drives visitors in line with what they did in the site, and then he determines which one leads to buying or dominant sign-ups in email. After site testing, the site controller could then update his website to employ the well-generated or valuable content.

On the other hand, A/B testing refers to a site administrator runs an examination by constructing varied types of a page with separate URLs. Thus, when a user attempts in accessing the authentic URL, the administrator redirects some of which to each variant, and then make a comparison in users’ attitude to discern which type of page efficiently works. Speaking of multivariate test is done when a webmaster utilizes software in altering the parts of the website where in testing can involve multiple pages thus the software delivers variations of each parts in variant combinations to users and it also analyze which combinations are effective to utilize. It should be noted, however, that, on the latter, involved a single url, and that deviations included strategically on the page.

Meanwhile, the main issue still is not on how the process of testing is going through or how a webmaster does it, but on how such usual venture affects or changes the site’s ranking. It was explicit in the words of Moskwa that the said testing is really a factor in the Google’s ranking. Nevertheless, there was no further explanation on why is that matter happens so; what the analyst team had just provided are the following helpful guidelines for employing a safe and effective site testing with lesser effect on the search function:

NO CLOAKING• Cloaking Not Allowed – as defined by Google, cloaking is about showing different piece of content to people and Googlebot which is violative of the guidelines of Google even if a webmaster is not doing any test. Hence, it is necessary for you to ensure that you’re not making a decision as to whether or not to operate the test or what variant of the given content to be used. One example of which is when you always serve real copy when you notice the Googlebot. Thus, always remember that infringement is always contrary to the regulations of the Google, otherwise, your site will become out of its “sight”.

Testing• Only run the experiment when necessary– it has been noted by the said analyst that the length of time required for a valid test varies depending on circumstances such as the rates in conversion, and how much traffic a website acquires; thus, an excellent testing tool could signal you when you’ve compiled sufficient data to draw a logical conclusion. And once you’re done with the initial process, you should then update the site with the needed content variations and reject the test elements immediately such as the markup and testing scripts. That is so because if the Google discovers that a website is utilizing an experiment for a long time, they may be misled to consider it as a deceptive strategy against the search engine and, consequently, your site might get penalized. Hence, you should avoid serving a content variant to a huge number of users.

REL="CANONICAL"• Use rel=“canonical” – if you’re doing A/B testing having varied URLs, simply employ the rel=“canonical” link attribute to different URLs in order to signify that its original is the desired version. The Google recommends using it than using a meta tag that is a noindex, according to the said analyst, it nearly matches with the webmaster’s intent in such situation. For instance, as extracted from Moskwa’s article, when you are testing variations of the homepage, you do not desire search engines in disregarding indexing of your homepage where in you only hope them in understanding that the tests are relative deviations on the authentic URL and should be classified as such and with the first URL as the canonical. On the other hand, utilizing noindex instead of rel=“canonical” in the said situation may have unexpected consequences; for instance, when Google selects one variant as the canonical, the original URL might fall from indexing and it could be treated by the engine as a recopy.

302 Redirect• Using 302s, not 301s – when an A/B testing redirecting users from original URL to a variant, then you must use a temporary or 302 redirect, and not the permanent or 301 redirect. This simply implies that the search engine is being signaled that such redirection is just temporary which means that it would only be set when you are doing the examination, in order that search engines must hold original or real URL in the index instead of the replacement redirection.

It was further positively noted by the analyst team of Google that by following the said guidelines, you can be assured that in testing your website, there can only be minimum or even no effect on your website in search engine results or ranking. Yet, it was also disclosed that it would not matter if the Googlebot indexes or crawls some of variations in the content while doing the process; but such still depends on what type is being examined. Furthermore, light transformations in color, size, and placement of image can possibly have significant impact on visitor’s interactions with the page of your site, but usually, will have no effect on the search result of that page. Nonetheless, if the Google consistently crawls your site to analyze your experiment, then it may be impartially indexing the updates to the site soon after the end of the whole process.

Al Gomez

SEO Consultant, Online Marketer & Blogger, Web Developer & DLINKERS Founder.

“Chose a job you love, and you will never have to work a day in your life.”