Even before the age of automation, good businesses monitored customer responses. It might have taken the form of a friendly shopfloor chat between a board member and a worker in the front line, or it might have been a more formal customer survey.

The arrival of information technology in the 1960s provided the opportunity to gather and analyse a wide range of data about customers. Tracking sales trends and buying patterns, for example, enabled businesses to gain greater insights into what customers wanted and re-organise their operations to give greater satisfaction.

On the surface, the growing use of the web as an important sales channel is more of the same. But in reality the web adds three new dimensions to customer data. Firstly, there is a lot more of it and, secondly, it is dynamic rather than static compared to the transactional data from established operational applications.

Finally, and most important of all, it gives an insight into the way customers behave. It is possible, for example, to monitor and record every detail of a potential customer's visit to a website. The buttons they choose to click on, the pages they linger on and their response to the design and navigation aids on the site are all recordable and potentially useful in revealing customer behaviour.

The combination of large volumes of data, its real time nature, and the potential for gaining greater customer insight demands new approaches both to gathering and analysing customer data. Established data analysis specialists have upgraded their software tools to cope with the higher volumes and the dynamic nature of web-generated customer data. At the same time a new breed of so-called web mining products has emerged to challenge the established suppliers.

"The web means a massive increase in data. The problem is to identify the useful and actionable data," says Laurence Shaw, European chief operating officer at consultant Headstrong.

"You can capture data on who hits the site and the pages they view. But the real value comes from tracking a whole visit - or preferably multiple visits to a site. Then you can start to build up a picture of customers and their preferences."

Mr Shaw adds that businesses can use "cookies" - a technology that can identify a customer as a frequent visitor - and personalisation of the site to gather such data. "You can then tell whether they are a new customer, a previous visitor or a returning loyal customer and treat them accordingly."

It is this ability to tailor a site visit to particular customers that brings out the real value of site monitoring.

Berni Simmons, UK country manager for data analysis tools developer SPSS, says that data from websites must also be combined with other sources to get the full picture.

He adds: "We have been in data mining since the 1960s and the basic premise of data analysis is still the same. But the web brings new and different data - especially to do with sequences of events.

"If you consolidate this with the traditional data such as customer details and transaction records, you can start to relate it to organisational activities."

Sandy Carter, IBM vice president of marketing for its Websphere product, says that making the connection between customer behaviour and site organisation is especially important in the web environment.

"There is a lack of personal interaction on the web so you must use the business intelligence data you gather to keep track of customers," she says.

Ms Carter goes on to illustrate the point with the case of lingerie site Victoria's Secret: "When they first set up the site they expected it to attract women in the 20s and 30s and designed it accordingly. But the demographics showed that most visitors were men between 40 and 55. Clearly, they had to change the design to appeal to them too."

She adds that the primary purpose of gathering data is to turn visitors into buyers: "In business-to-consumer commerce the conversion rate is only around two to four per cent. In business-to-business, it is much higher, but still only 20 to 30 per cent. If you can raise this by only a few points you can increase sales significantly."

As with SPSS, IBM takes the holistic approach - pulling data together from many different sources to enable complete analysis. "You have to take a multi-channel approach because that is what buyers do," says Ed Harbour, director of Websphere commerce at IBM.

"Buyers might place an order in a kiosk and then track its progress on the web, for example. Our Merchant Reach technology gives you a single source linking all the data so you can view the customer as a single entity."

NetGenesis, a US-based specialist in what it calls "e-metrics", has worked with both IBM and SPSS in combining traditional data with web-generated data. Anne Estabrook, vice president of marketing at NetGenesis agrees that any monitoring exercise must take account of all data sources.

"You need to optimise the performance of the web within the context of the broader business picture so you can align it with the business drivers," she says. "The rich behavioural data you get from the web is not available in normal channels."

This, she warns, is not easy: "It is much more difficult and completely different because it is technically more dynamic, and on-line tracking is less mature than traditional data analysis."

The wealth of data about customers now available to businesses gives them an unprecedented opportunity to provide the highest quality - and profitable - service. But, as ever, enthusiasm for the technology should not overwhelm sound business judgement.

"As much as 90 per cent of the data collected is not used so business must focus on genuine business benefits that can come from monitoring their customers. Or they will end up with enormous databases that deliver nothing to the business," notes Mr Shaw of Headstrong.