Lies, damn lies, hits and views

Internet statistics offer a beguiling mixture of detail and imprecision. Endless data about individual web users' habits lies at our fingertips, yet considerable confusion awaits anyone who wants to know how "popular" their (let alone anyone else's) website is. What's the best measure to use, and how accurate is it? What is a "good," or even a respectable, result? And how can this be turned into revenue?

To cover basic ground briefly, the number of hits a website claims to receive tells you the number of requests its server gets for file downloads over a certain period of time. Every web-page can have many files—text, images, sounds—so this is usually a large number that doesn't accurately reflect popularity, especially on complex pages with many images and scripts. A page view is a slightly better stat, in that it tells you how many distinct pages are accessed from a server, counting each page once no matter how complex it may be in terms of files. Counting visits is still better, in that it attempts to count each individual visit to a site only once within a given time (a set "session" length), no matter how many pages are viewed by each visitor.

Best of all is a site's unique visitors, which records the number of individual internet users who access a site within a given time period (usually a day, week or month). So, for instance, when a site talks about its unique monthly visitors, each internet user logging onto the site—as uniquely identified by data swapped between their computer and the site in question—can count only once towards the total. For example, individual sites such as personal blogs may get only between zero and 50 unique visits each day; young blogs that have begun to develop a following may get up to 500 uniques a day; more mature blogs might hope for 1,000 or more; well-linked, well-established blogs may start reaching over 10,000; while internationally-read, opinion-forming ones can pass 50,000 or even 100,000.

Such statistics are only the beginning of the debate. There are, for a start, all sorts of problems with properly identifying visitors as unique—and then there are the equally pressing questions of quality and quantity. Take last year's heated debate over which "quality" UK newspaper had the most-visited website. Somewhat dubiously, the Telegraph claimed this honour in its advertising, despite the fact that, by most measures, it only ranks around third, behind the Guardian and the Times. The Advertising Standards Authority, however, upheld the Telegraph's claim on the basis that it could be claimed that more people had used the Telegraph site than any other paper's over a certain period of time, according to one particular statistical measure.

The problem is, of course, that there are a number of rival ratings systems in existence for measuring the stats on big websites, and each measures slightly different things. NielsenNetratings, comScore, HitWise and ABCe are four commonly used systems, all of which produce data based on samples of audiences rather than raw server statistics (seeing as they don't actually have server logs for the sites they measure). Of these, it was only HitWise—which measures size in terms of market share based on numbers of visits over the course of a month—that ranked the Telegraph top, and only then among UK users. And this ignored the fact that, while the Telegraph runs its entire web arm off a single domain, www.telegraph.co.uk, papers like the Guardian have a web presence at multiple domains.

At the end of the day, the people to whom such statistics really matter are advertisers; and what really matters to advertisers is a website's relationship with the products they're trying to sell. The number of times an ad is clicked on is, for instance, a far more valuable figure than any number of visitors; while the "stickiness" (how long people linger) of a page is more crucial than how many times it is seen. Certain words and google optimizations can pull in thousands of visitors to a page—but giving users something they'll linger over, return to, and possibly follow an advertising link from presents a rather different challenge. With online advertising models becoming steadily more sophisticated, and companies increasingly realising that having lots of users and making a website profitable are rather different propositions, don't expect anyone to take you seriously if you can't muster a sufficient contempt for the crudity of almost all the "old" standards…