• GE Capital


     

    Striking commercial gold with data mining at GE Capital

    “Our purpose is to drive the business – to use data mining and advanced analysis techniques to help identify risk, to retain customers and to target prospects .” That is how Garth Waldron, Director of Global Modelling at GE Capital Global Consumer Finance, explains the role of his Global Modelling team. What may sound like a bold statement appears even more radical when you recognise the size of the organization within which this team (currently a dozen strong) operates. Global Consumer Finance (GCF) is a subsidiary of GE Capital, the financial services arm of General Electric, and provides retailer finance schemes in 25 countries outside the USA. The UK operation alone employs more than 1,400 people and provides private label credit card services to Harrods, the Arcadia, House of Fraser, Laura Ashley, Kwikfit and Dixons.

    Data mining techniques may have a track record of marketing hype and project failures, but it is clear that GCF is committed to ensuring that the technology delivers tangible benefits. The Global Modelling team is largely comprised of high-calibre individuals with PhD qualifications in electrical engineering, operations research, economics and other quantitative subjects. Garth Waldron reports directly to the Senior Vice President, Technology, in Stamford, USA and was recruited from Citibank, where he served for eleven years as a Vice President, involved with insurance, retail marketing, credit card business and, latterly, with the development of intelligent systems.

    Although the team is based at Capital House, just outside Leeds, Garth Waldron emphasises that their role is global: “Data modelling can be used to support marketing, risk analysis and business operations. There is some understanding within the organization of the value of these techniques and my role is to raise levels of awareness and establish a reputation for what we are doing. Having started with projects in the UK, we now work with businesses throughout Europe, Asia and the Far East.” While initial meetings tend be held on site, in order to gain a clear understanding of the business needs, teleconferencing and video-conferencing facilities are used extensively to ensure effective communication throughout projects.

    The global modelling team has undertaken around thirty projects aimed at a variety of objectives, including increasing response rates to promotion; identifying potentially fraudulent activity; determining the best time to make collections telephone calls; targeting products via telemarketing operations; and clustering customers that may be suitable for cross-selling other finance products. Garth Waldron explains that team members require sensitivity: “The role is both persuasive and diplomatic. Effectively, we are saying to people that we can help them to find ways of running their business better than they are now. A champion/challenger approach allows us to construct data models that demonstrate how – without changing any existing processes – a business activity can be given a lift in performance.”

    The management disciplines required by the team are the same as those needed for most IT projects: Planning and identifying resources; arranging supply of good quality data; developing a suitable solution; testing the solution; and documenting the results. In practice, as Garth Waldron explains, data modelling projects at GCF come with their own particular challenges: “The majority of the projects have been initiated from within the team, rather than from requests by managers. In general, people are unfamiliar with the methods and the terminology. Crucially, the importance of good quality data means that around 80% of the total work associated with any one project may be invested in data preparation. Everything starts and ends with data.” He adds: “Since we are implementing solutions within an existing computer architecture, we may need to adapt the solution that has been developed. Above all, we need to minimise the disruption caused so that the business can keep running.”

    Garth Waldron believes that the concepts and potential of data mining are easily recognised by most business people. He even uses mining analogies; talking about “seams” and “nuggets” of data to help communicate the principles. “Most of all, people want to know that there are bottom line benefits – although there are different perspectives about what those are,” says Garth Waldron. “It is rather like asking a group of blind-folded individuals to describe an elephant. The one holding a tusk will feel something quite different to the one with a hand on the elephant’s side. Similarly, we need to ensure that we communicate relevant benefits to each of the different groups with whom we work.” He is quick to stress that, although the techniques are sophisticated and the team is highly qualified, data modelling is not an academic exercise. Testing of data is based on their commercial value, while academic work is restricted to ensuring that the team is up-to-date with available data mining tools.

    The Global Modelling team uses a variety of data tools, for example, to determine optimal values and to create decision trees (identifying variables that might explain an outcome). Amongst these is XpertRule Profiler – a software product that has been used virtually from the outset. The product had previously been used by one team member and was chosen after evaluation of 4-5 other options. The team was attracted by XpertRule Profiler’s graphical representation of results and ease of use and the product has been used in a range of projects, including the development of telemarketing models for insurance products and identifying potential prospects for cross-selling of loans to existing cardholders.

    In client/server mode, XpertRule Profiler avoids the need to extract data from the data source. All information analysis is performed by using SQL to query the data for information. Compared to downloading data records, this produces very little traffic between the client and the server and puts very little strain on the client. The Global Modelling team is using its own network – based on powerful Sun client and server machines – and aims to capitalise on GCF’s own IT infrastructure, which also utilises client/server architecture for its business delivery systems.

    Data mining techniques have been tested successfully in other parts of General Electric and Garth Waldron’s team is expected to expand as the benefits are recognised within GCF (the team is already set to grow by 50% within the next three months). Ultimately, it is expected that continued growth may result in satellite groups becoming established in other countries to place them closer the businesses and to allow a faster response.

    Garth Waldron firmly believes that the growth of the team will be a result of commercial benefits: “We are simply using the information we have to help us find potential business; to win new business; and to retain existing business. As far as I am aware, this gives us a unique position in the consumer financial services market and gives us a powerful competitive edge.” He adds: “This organization is highly data-driven. The investment in setting up and supporting the Global Modelling team is a reflection of this and will help to ensure that we provide the right information of the right quality to the right people.”

     

  • Try Now & Live Demos