On Wednesday Google unveiled a brilliant new tool, the Web Optimizer, at this year’s E-Metrics Summit. It puts the power of multivariate testing on an array of web metrics in an elegant and simple-to-use online tool. What was once a tedious process involving capturing web statistics, downloading log files, importing them in a statistics package, and performing complicated regression analysis has now been made simpler. Users will be able to more easily test multiple versions of a page and the effectiveness of individual elements on those pages to determine which combinations result in the highest conversions.
From the press release:
This multivariate landing page optimization tool enables marketers to test different ideas for variations of headlines, promotional copy, or images and provides easy-to-read graphs showing which variation resonated best with their site visitors….
By giving website owners the tools they need to improve their website content, Google is helping improve user experience on the internet as a whole.
By allowing site visitors to determine what content is most useful, as indicated by the highest conversion rate, we are removing the guesswork and trial-and-error experiments that used to be the norm for determining landing page optimization.
Presently, Web Optimizer is in a limited beta and those interested in taking it for a spin are asked to apply for the beta test. A fairly comprehensive overview of the tool is available and contains screen shots of a sample combination report and page section report.
Sample Combination Report
Sample Page Section Report
What Web Optimizer Will Allow Users To Do
Google claims that Web Optimizer will allow users to ‘test different ideas for variations of headlines, promotional copy, or images’. However, with a little imagination, insight, and trickery the service should also allow one to test versions of graphic designs elements, navigation menus, page layouts, optimal text sizes and line heights and more. Theoretically speaking there is no real limit to the kinds of variables that one can test using multivariate analysis or this tool.
What Web Optimizer Cannot Do
There is an important caveat to the above statement outlining what the tool will allow user to do: technical limitations, dynamic elements, and variables that are not discrete elements on a static page limit the utility and capacity of Google’s Web Optimizer. In other words, the tool will not work on all pages and will not work on certain dynamic elements within a web page. It will also not be able to track the user interaction across different pages and states. For example, it will not give any insight into how various aspects of a user interaction impacts conversion rates for multi-state or multi-page dynamic systems like shopping carts and applications.
In addition, Google is marketing Web Optimizer as a tool that the average person can implement and use. The official press release reads: “This tool does not require consulting or professional services to implement.” Those that do not have an analytical background or training will likely be limited to carrying-out simpler tests; despite what Google claims, professional help will still be needed for more in-depth research. Perhaps it is without coincidence, then, that a consulting firm specializing in multivariate testing is collaborating with Google on this venture.
What Does It All Mean?
I cannot understate the importance of this service for the greater internet community. It will give web site owners, administrators, developers, and designers a tool to conduct basic multivariate testing to optimize single landing pages. This is significant. However, Web Optimizer only scratches at the surface of a comprehensively optimal user experience, and user experience designers, analytical consultants, and web designers will certainly not become obsolete. In fact, this tool can make those professionals even more effective. We at Montparnas hope that this tool will be made available to everyone, and we hope to use it as an integral part of our services. I myself will be all too happy to scrap the process of importing CSV files and doing all the multiple regression analysis by hand!