The bank now routinely mounts a new customer campaign once every two weeks, seeing take-up rates of between three and five per cent - up from one per cent before - and has achieved new business worth an estimated 50 million Euros in one season alone.
Behind the gains is Extreme Data Mining technology from vendor KXEN. It has brought improvements in the accuracy and speed (and hence frequency) of the data analysis used to identify potential customers. Just as important it has made complex data mining much easier, allowing it to be used by many more bank employees than before.
The implementation of KXEN's technology followed a bank-wide 'fit for sales' programme begun in 2004 and aimed at bringing all sales and marketing activity under central control, together with a reform of direct marketing. As part of these changes a higher frequency of marketing promotions was vital, but that was something the bank's then data mining platform could simply not support.
Werner Widhalm is head of the customer knowledge management unit at BA-CA. "It was critical for us to be able to conduct 14-day marketing campaigns based on relevant customer data, but traditional data mining had been far too time-consuming and complex to allow a fortnightly programme to be established," he explains. "The self imposed mission became one to find a tool for data mining which accommodated a heightened demand for speed and precision."
Working alongside Widhalm on the project team was a management consultancy firm. It very quickly identified Extreme Data Mining as a perfect fit for the bank's needs, ran a test and within a few weeks KXEN's technology was chosen.
Speed and accuracy aside, it was the KXEN solution's ease of use even by non-specialists that was the most compelling factor in the bank's decision. Most other data mining tools require specialisation in mathematics or statistics before they can deliver meaningful results.
"Anyone who has a reasonable amount of experience in data analysis can quickly acquaint themselves with the software," says the bank's Erich Hrusa, responsible for technical architecture in customer knowledge management. "We also wanted a fast-working system to meet the demands."
Analytical models created in KXEN are automatically fed through the bank's scoring engine in batches weekly or monthly depending on the schema. "In a month we can now run a minimum of 20 analyses, something which before would have taken at least four months to process, thus improving our time to market" explains Erich Hrusa.
Specific applications of KXEN's Extreme Data Mining technology include prediction of propensity-to-buy, customer segmentation (cluster analysis) and retention analysis. Results from KXEN analyses are fed back to the data mart from where they go into the bank's Epiphany system where they inform marketing campaigns.
Such has been the success of the new KXEN system at BA-CA that staff no longer consider the old way of data mining - users developing regression models over long periods of time - to be viable. They believe the KXEN software itself holds the mathematical expertise required, with the expert adding the all important business knowledge that completes the process. "This practical approach works very well," says Hrusa.
As well as producing models at high speed the KXEN software also rapidly evaluates their quality. The end result is vastly increased speed and productivity, making modelling effectively a production line activity. "Our data mining models are far more industrialised in comparison to previous ones: this is the only way to manage the plethora of marketing activities now," he says.
But what about the success rate of predictions? Werner Widhalm again: "It is difficult to make an exact evaluation as there are generally about five to eight parallel marketing campaigns, which may be competing with each other. But we are looking at a success rate of target customer deals in the area of three to five per cent with KXEN. Before that, it was one per cent or less."
Also indicative is that data mining now supports around 20 per cent of new business at the bank, which added up to some 50 million Euro last spring. "Thanks to good data quality and more qualified information we achieve better results even though we approach fewer customers," says Widhalm.
There have been other benefits too. One example is in analysing customer churn. "We have over 1,500 attributes and patterns on old customers who have left, therefore the factors that might indicate a current customer about to churn would simply not be discernible to the naked eye," says Widhalm. Now if a customer displays certain behavioural patterns - such as termination of products or a decrease in volume - staff can spot it in time and take appropriate action.
Impressive though the results to date have been, BA-CA is already thinking of ways to improve them even further. One example is evaluation of methods where the results of relevant success measurements flow back into the model, together with incorporation and further analysis of sales from operations. Following best practice the bank also intends to extend and develop in the direction of centralised data processing, with a view to creating what it calls a ‘single point of truth’.
In the analytical area Werner Widhalm feels the goal should be to provide an increasing level of detail about customers and so gain still more information for data mining. "The more we know about our customers the better we can serve them," he says.
|< Prev||Next >|