Principal Scientist, Qatar Computing Research Institute, Qatar
Professor, College of Science and Engineering, Hamad bin Khalifa University, Qatar
Web search, web searching, search engines, information retrieval, information searching, information searching and retrieval, ecommerce, sponsored search, online marketing, online advertising, online branding, Google AdWords, PPC, Pay-per-click, keyword advertising, second screens, cross screens, dual screens, customer segmentation, audience segmentation, user segment, persona, personas, persona profiles, information science, library, web analytics, digital analytics .
Science wants numbers about people: The numbers are likely wrong
Science (as well as the industry, government, non-profits, and university sectors) want, use, and depend on numbers for situational awareness and then make decisions based on them. However, what if the numbers are wrong? Unfortunately, there is ample evidence that the numbers are often wrong for numbers concerning people and behaviors. As a case study, we ‘peel back the onion’ on the numbers from a large-scale web traffic study of 86 worldwide websites comparing two web analytics tools. One tool is a site-centric service, and one uses a triangulation of user, site, and network-centric methods. The resulting numbers for the three standard analytics metrics evaluated are statistically different. However, which set of numbers is ‘more’ right? A deductive analysis shows that both approaches are inaccurate and imprecise, even though advertising and other decisions worth hundreds of billions of US dollars annually are made using these numbers. The talk ends with recommendations to move forward for the reporting of numbers in this and related contexts.