The value of tracking customer behaviour is obvious to most businesses. It helps them improve customer interactions and identify possibilities for increasing sales. With this in mind, many vendors, and particularly those from the customer relationship management sector, have begun to promote the concept of the single view of the customer.

There are real benefits to unifying customer records. By understanding when multiple records relate to a single customer, suppliers can distil their marketing efforts, while potentially being able to offer more attractive deals.

But actually getting a single view can be a complex process. Frequently, customer records will be stored in departmental silos, says Richard Kellett, strategy manager at business intelligence vendor, SAS. "Everyone is willing to accept that having a single view of the customer is vital. But only as long as it's their view," he comments. "The biggest single question is who owns the data, and then who owns the definition of that data?"

The separation of data sources also introduces the likelihood of anomalies when trying to generate the ‘single view'. At financial services company Barclays the solution to this problem was to introduce matching software from BI vendor Trillium. "There's a lot of science to it, but it is also an art," explains Matt Harris, head of CRM at Barclays. "You get to the point where it gets very complicated, removing vowels from names, looking at percentage match rates. That's where the art comes in."

However some organisations do not believe that software tools are a worthwhile investment when attempting to verify customer data. Kate Brooks, manager of wholesale data management at Lloyds TSB estimates that over 10% of their addresses need to be corrected to support a single view of customers. Lloyds TSB operates three business units – asset finance, corporate banking and treasury – where it may have customer overlap. To limit the number of errors, it encourages staff to check the veracity of records at the departmental level. Address validation tools have proven to be "rubbish" says Brooks.

And while the notion of a single view of a customer may appear to be appealing, it is worth applying a ‘fit for purpose' test. If the purpose of distilling the information is to meet regulatory requirements, attaining 100% accuracy is imperative. But for marketing purposes, perhaps a lower rate is acceptable – and more justifiable.

While it is possible to check data retrospectively, it is far easier to get the correct data up-front – and banks get a head start since they impose such stringent identification measures when new accounts are opened. While dirty data makes the quest for a single view of the customer difficult, the level of data cleaning needs to be tailored to the return.