UK web stats 'notoriously inaccurate'

The world of accounting has moved on

By BBC News Online's Ivan Noble

Three quarters of UK websites do not count their users properly, the European head of an international research company says.

Any business that is going to survive is going to survive only if they understand their consumer

Martin Filz,
RedSheriff

Martin Filz of RedSheriff says that the way most commercial websites measure their traffic is "notoriously inaccurate".

"They are not measuring what is delivered. They are measuring what is requested," he told BBC News Online.

One might think that the two things should be the same, but a range of ways in which the web works means that companies often end up overstating figures like page impressions and unique users.

Unwitting inaccuracy

But it is not that the UK industry is plagued by fraudsters; more that firms do not understand the technical complexities of audience measurement and so misrepresent their statistics unknowingly, according to Mr Filz.

Of around 100 UK firms surveyed in 2000, he found that 75% were inaccurately reporting their figures, mainly because they were using server logs to count their users or they had misunderstood the way their auditing software worked.

When a web user looks at a web page, the individual elements of the page are delivered to the user by a server, which then records details of which page components were requested and whereabouts on the internet they were sent.

Statistical tools process these server logs to try to deduce important information like how many people use the site, how many pages they look at, and for how long.

Network trickery

Problems arise because internet service providers use a variety of tricks to speed up their service to users and make use of scarce network resources.

They often store their own copies of popular web pages so that they can deliver them to their users more quickly, a process known as cacheing.

When this happens, the server only records one request for the page, even though several hundred internet users may be looking at that single copy.

And while auditing software looks at network addresses known as IP addresses to try to work out how many individual users a site has, internet service providers juggle these addresses around between users because there are not enough of them.

The networking technology which underlies the internet was designed long before anyone envisaged it becoming as popular as it is today.

Better accounting

Martin Filz said that businesses could get round such tricks by using software which measured what happened at the user's end of the browsing transaction, rather than at the server end.

This involves using a range of techniques to embed small pieces of code into web pages.

The code runs on the user's computer and reports back to the website.

Using three separate tools - Java applets, common gateway interface (CGI) scripts and cookies - covers a good range of different browser and security setups and provides much better information, he said.

Privacy concerns

Many internet users are concerned about such techniques, because they worry that they could compromise their privacy.

The techniques are indeed powerful, but maintaining privacy depends on the way that they are used.

Martin Filz said that his company had a clear policy on privacy:

"We don't gather any personal information. We don't keep track of which other web sites you visit," he said, adding that he wanted to see a process of education explaining how such tools could provide useful information without compromising privacy.

Uphill task?

It might seem like an uphill task to persuade internet companies to use new accounting methods, which then tell them that they have fewer users than they thought.

But having clear information about users is in a company's own interest, Mr Filz said:

"Any business that is going to survive is going to survive only if they understand their consumers. And that means knowing how many there are, what they're doing and what kind of people they are."