Diversified companies, like wide lapels and shoulder-length hair, have been out of fashion since the 1970s. With one or two notable exceptions, such as Jack Welch's General Electric, the capital markets prefer simplicity and focus.

This may be about to change. Enthusiasts believe that the ability of companies to collect, store and analyse customer data may yet provide a new justification for the existence of diversified, multi-divisional corporations. And sceptics argue just as fervently that companies must overcome huge obstacles for diversification to work.

In theory, companies should diversify only if they have the "strategic assets" to win in a new line of business. In the past, these assets might have ranged from engineering excellence to distribution clout. Today, the question is whether the ownership of customer information is also becoming an advantage. "If customer information is a strategic asset, you can see that there may be an argument for managing a wider range of businesses within the same entity," observes James Lattin, associate professor of marketing and management science at Stanford University's graduate school of business.

Importantly, you can put a value on the relationship with each customer. The institution can work out which are profitable. This analysis can generate eye-opening results. In her book Customers.com, Patricia Setbold, a US technology consultant, recounts the story of a bank which discovered that 20 per cent of its customers accounted for more than 100 per cent of profits. The others were destroying value.

At the same time, financial services deregulation has made it easier for financial holding companies to use this data to cross-market between products. The bancassurance model is the direct result. Long-term savings products are sold alongside short-term deposits, with intensive use of customer data at the core of the strategy.

One example is the acquisition last year by Charles Schwab, the San Francisco-based stockbroker, of US Trust, an upmarket fund manager. While Schwab has its roots in low-margin discount broking, customer analysis revealed that a high proportion of its clients had grown rich on the back of the long-running equities bull market.

By purchasing US Trust, Schwab hopes to deepen its "share of wallet" by offering financial advice and wealth management alongside its core business. Again, data-mining is an important part of the strategic jigsaw. Two complementary trends suggest that other industries may be able to pursue similar multi-divisional strategies.

First, spending on customer data systems, bundled broadly under the title of customer relationship management (CRM) applications, was one of the big items in the information technology spending binge of the late-1990s. This should equip more companies with the tools required to individuate customers.

Second, more companies are defining themselves by reference to their customers rather than to their products. This often involves the introduction of a "customer-facing" management structure, with divisions set up to answer the needs of particular market segments.

Professor Lattin cites the example of Hewlett-Packard under chief executive Carly Fiorina. The Silicon Valley pioneer has overlaid its traditional product-based divisions with a system of customer-facing management. Another Silicon Valley example is the "vertical service providers", technology companies which aim to supply the IT requirements of an industry segment rather than concentrating on a technical specialism such as databases or e-commerce.

It is unclear whether this emphasis on definition-by-customers is part of a secular trend or just a management fad. Either way, it is prevalent. David Montgomery, professor of strategic marketing at Stanford, says that customer value has replaced cost control "as the guiding principle of management". If he is right, expect more companies to maximise the value of CRM by cross-selling products and services.

CRM systems of the kind sold by vendors such as Siebel, SAP and E.piphany promise to help take the guesswork out of selling new products and services to existing customers. Roger Siboni, E.piphany's chief executive, argues that advances in artificial intelligence (AI) and data processing speed make it possible for companies to target their cross-selling efforts more effectively.

For example, call centre operators can be given instant analysis and a suggested course of action for each caller: "suggest product X", "cross-sell product Y", or "back off!"

Mr Siboni admits that such AI-based systems will never be foolproof. None the less, even an incremental gain in the rate of cross-selling can have a big impact on profits. After all, selling new products to existing customers avoids the need to spend money on advertising, marketing, administration and all the other elements of "customer acquisition".

The prospect is intriguing. But do not expect customer-facing "new conglomerates" to evolve rapidly, or to resemble the corporate dinosaurs of old. Several factors will hamper the evolutionary process-and, say sceptics, will halt it altogether.

For a start, despite the bold claims made for CRM applications, they are difficult to implement. Research last year in Europe by Gartner Group found that the failure rate for CRM projects was close to 50 per cent. This may explain why spending on CRM technology has slowed to a crawl this year as companies have cut IT budgets.

Even when successful implementation has been achieved, the technology is a long way from achieving a perfect understanding of the fickle behaviour of customers. Rates of cross-selling may improve but, as Professor Lattin says, "there is no magic bullet".

James Schrager, professor of strategic management at Chicago University's graduate school of business, is an arch-sceptic. "Data-mining is nothing new," he says. "The direct mail industry has been at it for years with only limited success. All my experience tells me that consumer behaviour is so complicated and inconsistent that no amount of data and analysis will deliver significantly higher returns."

Moreover, even companies that define themselves in relation to their customers will be wary of straying too far from what investors regard as their core business. The continued prejudice of capital markets against diversified, multi-divisional companies remains a powerful discipline.

This suggests that managers will explore cross-selling opportunities using alliances, partnership programmes and joint ventures rather than full mergers and acquisitions.

The drawback with these "virtual" conglomerates, however, is that customer data needs to be shared between companies. This may not be an issue if customer information is just another part of the marketing tool kit. But if the data starts to be seen as a strategic asset, the keys to the CRM servers will not be handed over so readily.