Data mining to infuse decision-making capabilities throughout enterprise CRM solutions

Data mining makes sense of mountains of data, enabling better, faster and more accurate decision-making to solve critical business issues. Its value is increasingly being recognised by a wide variety of companies that are using the power of business intelligence to boost the bottom line and gain monetary or strategic advantage.

A Meta Group report predicts that data mining models will become easier to integrate, and less expensive to deploy in the business process during 2002 and 2003.

According to the report, during 2002-03, data mining will provide large and mid-sized enterprises with detailed classifications to group customers into a variety of easily actionable segments. By 2004-05, such analytics will have infused `actionable` decision-making capabilities throughout the touch points of enterprise CRM solutions. (Top Five data mining trends for 2002-03, by Aaron Zornes, first published by Meta Group on 13 December 2001)

The report projects that adoption of enterprise data mining solutions will be accelerated by more predictive models, better data mining models, more cost-effective modelling, evolving data mining standards and integration within RDBMS servers.

Bruce Jones, sales support manager at SAS Institute, agrees: "More power and versatility in new high-end data mining tools enable a rapid modelling environment which can solve multiple business problems almost simultaneously. Powerful statistical techniques allow users to combine single models into meta-models. This gives them even more options to make the best business-driven decisions."

As the average skill level of data mining users increases, and with the ever-growing demand for better and more accurate results, the quality of data mining models is becoming a prime focus in data mining development.

"New and improved algorithms, better model evaluation, higher accuracy levels, and shorter response times, are only a few of the factors that will streamline the further development of data mining," says Jones.

To fully reap the maximum rewards of data mining, models have to be more effectively implemented in business processes.

"Data mining developers are currently focusing on improving integration with operational systems, such as CRM," says Jones. "Business users want front- and back-office systems tightly coupled. This would contribute enormously to decreasing the `total cost per model` and have a major impact on the effectiveness of overall CRM strategies."

Data mining standards are evolving, and a current trend is to create multi-model, vendor-independent platforms. These will allow users to develop a model using one vendor`s application, and then use other applications for visualisation, analysis or evaluation.

"This will allow disparate data mining and business intelligence applications to be used together, without the need for manual intervention," he says.

"Another goal is to integrate models in real-time operational systems, thus reducing the cost of deploying new models or updating old ones."

Jones says that integration of data mining with RDBMS servers is the subject of much debate.

"Leading market players agree that this will eventually happen," he says. "However, even though technology is available, there are major obstacles in its path.

"Firstly, concurrency problems exist, which can only be overcome if the database is separated from the operational environment. Secondly, it is clear that successful integration is fully dependent on parallelism.

"The Meta Group predicts that by 2005, parallelism will be a native feature of all high-end data mining capabilities. When this will become the norm is another question - given the high cost of parallel architectures."

Jones concludes that the development in data mining is ultimately dictated by its impact on the heartbeat of businesses.