Brenda Dietrich is a research fellow within the mathematics department of IBM's research division. Dietrich attended IBM's launch of Cognos and spoke with eWEEK senior writer Renee Boucher Ferguson about her department's plans to work with Cognos, and what might be accomplished by applying mathematical techniques to traditional business intelligence.
What is the focus of the mathematics department at IBM?
The IBM research division has had a mathematics department for more than 50 years. Our role is to work with clients on everything from supply chain and product design to demand forecasting to [building] models to better understand and predict customer behavior.
How does that translate in terms of the work you do on a daily basis?
We observe a business situation-with supply chain I spent a lot of time in IBM manufacturing, watching how it's done. We decide how decisions are made then use mathematics to figure out the best decisions. We instantiate this in software-large pieces of code-that are generally hooked up to ERP [enterprise resource planning] or CRM [customer relationship management] and the algorithm sucks data out of the system and makes recommendations.
How accurate are the recommendations?
Accuracy is a function of data quality, always. Data mining is statistics based and it's very accurate when it's in something that is approaching [stability]; you can't predict something if it hasn't occurred, if it hasn't shown up. In general [the predictions] are more accurate than the processes they replace.
How predictable are the scenarios in Cognos' world of business intelligence and performance management?
We're hoping to work that out with the Cognos team-the acquisition is only seven days old-and some customers to understand the problems that are at the cutting edge of what Cognos is able to address, [things like] dash-boarding or alerts, to put more predictive modeling into and to project [the] future.
Why is Cognos having such a big impact on your research?
Part of it is this area we worked in [Information On Demand] is growing really rapidly. It started very small but it really does require a full infrastructure to be more effective. Now we're in a market where customers have historical data. They've invested in ERP and as a by-product of investing in ERP you get trace data and this is a gold mine for doing modeling. Ten years ago it didn't exist.