In most companies customer-side data, such as complaints and comments, or supplier-side data, such as records of service delivery, are rarely matched with customer value and transaction data. If matching is carried out, it is usually on a batch basis, and can only be managed rather crudely. Only a few companies are experimenting with, for example, call records, to match the company's view of how well it dealt with the customer, how the customer perceived they were managed and what they thought about it, and subsequent customer behaviour.

In a project one of the authors is involved in, the aim is real-time integration between customer service, transactions and response data - and this means action, not analysis. The idea is that rather than wait for the customer to contact the company, the company will initiate the contact. So, when a customer gets into contact with the organization, in any channel, for any reason, the organization looks at its purchasing or response data and says, ' I see you bought X from us for the first time last week, did you like it?' Or, 'I see you used to buy Y from us but you've stopped - can you tell us why?' Or, 'I see you're buying less often from us - can you tell us why?' Or, 'You never respond to our direct mail or e-mails - do you actually read them?' Of course, this would be permission-based, so the first question might be, 'Every now and then, when you're in touch with us, we'd like to check that you're happy with us, so would you mind if we....'

Once the programme is going, the company will need to bring customer service, transactions and response data together quickly so it does not ask about ancient history (or as the customer might put it '...that was last week, things are different this week'). If the customer is in contact with the company several times a week (for example in a counter-service or Web situation), it will no longer be able to rely on batch analysis, but will need to set up automated processes for bringing together its marketing, sales and service data and actions. This approach is already visible in some companies that use only the Web and e-mail to manage their customers. Doing it in call centres, branch outlets, kiosks, ATMs and the like requires a big step forward.

The technology for doing all this is, as usual, there (though not very often deployed). The business case for doing it (and to whom) needs a long, hard look, of course. Here, it is important to apply classic direct marketing disciplines of measurement and testing, combined with good research. The latter should be used to identify the things that customers say really matter to them - the service that makes them buy more or less, or stop buying entirely, as well as the things they notice (what they say influences their behaviour and what actually does influence it may not be the same).

It is also important to take into account the leads and lags involved. Much academic and business research shows that few customers change their behaviour after one or two good or bad service episodes. For example, loyal and/or previously very satisfied customers tend to forgive bad service, while it may take several episodes of excellent service to make a customer respond really positively to it. However, some customers do respond quickly, so we need to analyse not just averages but distributions. Time must also be taken into account - how long it takes for customers to perceive improved treatment and when they are next likely to buy. In another project in a different sector from the one mentioned above, a slightly improved service treatment that is declared to customers and properly fulfilled has lifted response rates to mailings and subsequent purchase rates.