It was a bad day for Britan nica.com's marketing team. The staff at the online arm of Encyclopaedia Britannica had spent weeks negotiating a sponsorship deal that promised to bring in tens of thousands of dollars in vital revenue.

Then they hit a stumbling block - a key set of figures. During the negotiations Britannica.com claimed it was attracting several million visitors a month, making it one of the most popular web properties in their sector and an ideal target for sponsors.

But the potential sponsor - search website AskJeeves.com - had come across a different set of numbers from Nielsen/NetRatings, one of the leading companies in web traffic counting and analysis.

The Nielsen/NetRatings figures counted only a fraction of the visitors that Britannica.com claimed was clicking into its website every month. AskJeeves was not impressed and negotiated a much smaller deal.

"They didn't really believe our figures," says Taikwan Chin, Britannica.com's marketing and strategy manager, "The Nielsen figures were so different."

She is not alone in her frustration. Online businesses across the sectors are complaining about glaring differences between the visitor numbers they have logged on their own computers and the figures produced by "official" web surveys.

There is more at stake than pride and big numbers. Web visitor figures have become a key metric of success or failure in the increasingly pressured internet economy.

A rise or fall in audience numbers can have a huge impact on staff morale, market valuations and, most importantly in today's cash- strapped world, advertising and sponsorship revenue. Wild variations in audience numbers risk scaring off those advertisers and sponsors all together.

"Big online advertisers - the Procter & Gambles of this world - really need the internet to become more reliable (if they are) to feel comfortable investing large sums online," says Dakota Sullivan, vice president of marketing at Looksmart.com, the internet search site, which has its own complaints about the measurement of web traffic.

Many advertisers, says Mr Sullivan, just fail to understand the reasons behind the differences between the varying statistics. Despite the central significance of the site visitors, the industry had yet to settle on a standard way of counting them.

Nielsen/NetRatings and competitors, such as Jupiter Media Metrix, rely on panels of volunteers for their data. The companies fit devices onto volunteers' computers that monitor where they go on the web. Data from that panel is then extrapolated to represent an entire country, or even continent.

Supporters

At the other end of the spectrum, websites can choose to bypass the Netratings of this world and collect their own data from their own computers. Webmasters can look at the servers that power their homepages and count the visitors as they come in. Both methods have their supporters.

Fans of panel-measurement point out that they count people, not computers. They like to list the ways that computers can be confused. Hundreds of workers from an office might log on to a website in a week. But, because they all use the same corporate network, that website might think it was just a handful of visitors, coming in over and over again. Machines might also miscount the millions of people that get online through AOL who often share common network addresses.

So-called dumb computers can also log a single person many times if he or she gets to a website using many different PCs - perhaps at work, at home or the library. Servers can be tricked by 'spider' programs that automatically crawl round the web visiting sites 24 hours a day, notching up hundreds of visits.

Andy Fessel of Media Metrix estimates that server logs can end up counting between two and five times the number of unique users that his systems would recognise. "The simplest explanation for the difference is that Media Metrix and server log files measure very different activity and consequently report different metrics," he adds

At the other extreme, fans of computer logs boast about the power of their raw data. The information arrives in real time, minute by minute, as visitors log on and leave. Experienced web watchers can gauge how long visitors spent on an individual page, how they got there, and how often they returned. Log-file enthusiasts also like to point out the shortcomings of the panel crowd.

Limited panels of volunteers do a fair job of measuring the most popular sites online - the Yahoo!s and Amazon.coms of the internet universe. But they struggle to collect a statistically significant sample of visitors for niche sites - the smaller online businesses and personal homepages that crowd the internet.

Few of the panel companies have a significant presence outside the US and the more obvious European capitals. They might fall short in estimating a site's audience from emerging regions including parts of Asia and the Middle East.

Whatever the arguments about competing web counting technologies, there is little doubt that the demand for a reliable standard of measurement remains.

A study by Forrester Research sees the total web measurement market growing from $100m in 1999 to $300m this year and $500m in 2002. That assumes continuing stratospheric growth in web traffic seen in the late 1990s and early 2000s, not necessarily a given in today's uncertain climate.

A host of smaller companies have sprung up to take advantage of the opportunity. Many offer sophisticated analyses of log files, using the data to recreate every move of a visitor on the site.

UK-based Clickstream Technologies, for one, promises to track how long a visitor spends on a page, even if he or she is looking at a stored version of that page while they are offline. "The new technology comes amid growing concern over inflated website visitor figures among both marketeers and advertisers," it says.

Many websites are reacting to these concerns with a belt and braces approach of offering advertisers a variety of methods to gauge their web traffic. Theonion.com, the satirical news website, boasts 4m unique visitors a month. It backs it up with data from Nielsen//Netratings and I/Pro, and independent operation that, in effect, audits Theonion's own server files.

Online businesses are also choosing to rely on the old-fashioned power of persuasion to back up their various forms of statistics. Britannica.com's marketing staff eventually managed to persuade AskJeeves to renegotiate its sponsorship by showing how much traffic its millions of visitors were pushing through to the search site. The deal was signed, but Ms Chin knows AskJeeves will not be the last site to doubt her figures.

"It's generally a never ending issue," she adds. "It's always being brought up. All of the sales force and everybody in marketing knows the difference between our figures and Nielsen's numbers. They have to explain it every time."