Will we ever really get to grips with the word 'loyalty', or would it be better to dispense with it altogether?

The big idea behind loyalty - as first put forward by pioneers like Bain & Co's Fred Reichheld - is that loyal customers are more profitable. Keeping existing customers is cheaper than finding new ones. It costs less money to serve loyal customers. They provide the best opportunities to cross-sell and up-sell.

But as companies experiment with loyalty marketing, doubts are coming to the surface. One of the earliest nay-sayers was Professor Andrew Ehrenburg of South Bank University, whose long experience with purchasing data led him to argue that often the most 'loyal' customers are the least profitable.

In most categories the big spenders tend also to be repertoire buyers. It's the small spenders who buy the same brand again and again - but only very infrequently. This 100% loyal customer is worth hardly anything compared to the promiscuous 'flitter'. Conclusion: the 80/20 rule rules. Don't go for loyalty per se, go for the 20% of customers who really matter, no matter how 'loyal' they might be.

Ehrenburg's scepticism was matched by experienced business-to-business marketers who quickly pointed out that the most financially significant 'loyal' accounts are often the least profitable. They're the product of big, powerful customers muscling their way to such good deals that there's hardly any margin left for the seller.

More recently, McKinsey research in the US underlines the need to distinguish between three types of loyals. They are 'deliberatives' who rationally re-choose their current provider because it offers the best deal; 'emotives' who feel a special affinity for the brand; and 'inertials' who can't be bothered to switch. Clearly, each segment needs to be treated very differently.

The McKinsey research also suggests that people who spend less on a brand have a greater financial impact than people who actually defect. Over a one-year period, for example, one retail bank saw deposits fall by only 3% as a result of defections, compared to a 24% fall as a result of reduced balances.

These 'downward migrators' fell into distinct camps. There were 'lifestyle' downward migrators, whose changing lifestyles meant the brand no longer fitted as it once did; deliberatives, who had found a better deal elsewhere; and the dissatisfied, who had a particular complaint. Again, each group requires a very different management response.

Further research by Werner Reinatz, an assistant professor of marketing at Insead and V. Kumar of the Connecticut School of Business Administration (reported in Harvard Business Review) throws more cold water on loyalty. Having studied a hi-tech corporate services provider, a mass grocer, a financial services company and mail-order company, they conclude there is "no evidence to suggest that customers who purchase steadily from a company over time are necessarily cheaper to serve, less price sensitive, or particularly effective at bringing in new business".

In none of the companies were long-standing customers consistently cheaper to manage than short-term customers. In fact, because one-off transactions tend to be cheap to conduct while ongoing relationships are expensive to manage, the opposite was often true.

What's more, long-term customers tend to become more knowledgeable about their suppliers, and therefore more able to extract better value from them. They also "strongly resent" any attempts by the company to profit from their loyalty. Indeed, they often "expect something in return" for their loyalty, the cost of which often cancels out its original financial benefit.

If we add all this together, what do we get? First, the common assumption that increased loyalty equals increased life time value is questionable. Second, the correlation between so-called loyalty schemes and improved customer loyalty is suspect. Another conclusion of the Reinatz research is that 'attitudinal' loyalty is a far more important driver of word-of-mouth recommendation than the behavioural loyalty that's influenced by points-for-purchases schemes.

Does that mean we should abandon the whole loyalty concept? Is a combination of good old-fashioned brand building (which changes attitudes) and promotions (which change behaviour) all we need? No. The trouble with both these approaches is that they lie at opposite ends of the same spectrum. They're united in treating the customer as the passive target of one-way stimulus-response marketing programmes.

With 'loyalty' a new - and very different concept - is struggling to put its head above the parapet. It's about mutual accommodation for mutual benefit.

The difference between Tesco's ClubCard and many points-for-purchases loyalty schemes, for instance, is that Tesco uses the information it generates to change the offers it presents to its customers. The implicit 'deal' is that if you let us collect more information about you, we'll give you better value for money (points) and better service - while, at the same time, cutting our own costs and boosting our revenues.

Likewise, under good CRM programmes the message is: if we give you more say over how you interact with us and bombard you with less spam but more timely, relevant messages, then perhaps you will spend more time and money with us. So we both benefit.

Right from the start 'loyalty' was a horribly misleading word. What's struggling to emerge here is not loyalty but mutuality: negotiating new types of win-win with customers. Dispense with that at your peril.