Lloyds TSB - Data mining to exploit transactional data

Lloyds TSB is a leading UK-based financial services group, which was created in 1995 following the merger of the TSB Group and the Lloyds Bank Group. Its businesses provide a wide range of banking and financial services in the UK and overseas, principally through branches of the Lloyds TSB Bank and its wholly owned subsidiaries, Cheltenham & Gloucester and Lloyds TSB Scotland.

The Challenge

Lloyds TSB had installed a data warehouse, which captured all its customer’s financial transactions across each of its delivery channels. There were reams of dynamic data now available to the bank, which they wanted to be able to exploit in managing their customer relationships. Particularly they wanted identify:

	Cross-sell opportunities
	Customer servicing cost
	Customer value
	Customer churn
	Cost effectiveness of sales and distribution channels
	New business opportunities

CACI were chosen to mine this complex and transitory data to try to exploit the information buried within it.

The Solution

Transactional data was mined initially over monthly periods, to explore the short term characteristics of customer transaction profiles, in terms of their channel usage, transaction types & volumes and account status changes. Long term transaction dynamics were then assessed over a twelve-month period to investigate how these short term transaction profiles change over time.

The data was aggregated by transaction type, delivery channel and account balance history, so that different business issues could be examined.

An unsupervised neural network was used for the mining process, which allowed the data to be clustered according to physical similarity, rather than any predefined rules or conceptions. For each of the clusters detected the typical transaction profile was established. Individual clusters were then scored with additional attributes which were of interest to the business, such as churn rates, sales levels and recruitment rates.

The Results and Benefits

A number of interesting & potentially lucrative findings came out of the data mining analysis, including the following:

	Higher value customers were seven times more likely to use the Internet or telephone banking, partly due to their higher level of rural living. This resulted in a review of retail locations serving these groups as well as prompting changes in the marketing channels used to address these customers.
	High cost customers were shown to use branch outlets frequently for tasks that did not require their physical presence. Strategies were adopted to provide alternative, more efficient channels for these customers to use, if so desired.
	A significant opportunity for increasing investment business was identified based on a small segment of the customer base who were shown to be active investors.
	A subset of customers was identified that appeared to have the same opening and closing balance each month despite making transactions. This anomaly was traced to a problem in the data warehouse feed that had not been spotted before, hence the data mining analysis also fulfilled a vital additional validation role for the new warehouse.
	In addition, some suspicious activity was also identified by the mining process which was passed on to the banks internal security team.
	The data mining exercise has since been repeated on the Banks relatively static customer information database where extensive scoring in terms of risk and value has been carried out. CACI have also developed a ‘super-clustering’ method to combine the data mining results from this database and the transaction data warehouse. This has enabled the large wealth of information these two sources contain to be viewed efficiently & effectively.

To find out more about our work with Lloyds TSB either call 020 7602 6000 or email info@caci.co.uk