Doing the BI math
IBM researcher Brenda Dietrich discusses how math enhances
the effectiveness of Information On Demand and business intelligence
Brenda dietrich is a algorithm sucks data out of the impact on your research?
research fellow within the system and makes recommenda- Part of it is this area we worked in
mathematics department of tions. [Information On Demand] is grow-IBM’s research division. Dietrich ing really rapidly. It started very
attended IBM’s launch of Cognos How accurate are the recommenda- small, but it really does require a
and spoke with e Week senior writer tions? full infrastructure to be more effec-Renee Boucher Fergu- Accuracy is a func- tive.
son about her depart- tion of data quality, Now we’re in a market where cus-ment’s plans to work always. Data mining tomers have historicaldata. They’ve
with Cognos and what is statistics-based and invested in ERP, and as a byproduct
might be accomplished it’s very accurate when of investing in ERP, you get trace
by applyingmathemati- it’s in something that data and this is a gold mine for
cal techniques to tra- is approaching [stabil- doing modeling. Ten years ago it
ditional business intel- ity]; you can’t predict didn’t exist.
ligence. something if it hasn’t
occurred, if it hasn’t Will your research with Cognos result
What is the focus of the shown up. In general, in actual products?
mathematics department Dietrich: IBM has had a math [the predictions] are Probably pieces of an actual product.
at IBM? department for 50 years. more accurate than the I don’t think we will be likely to have
T h e I B M r e s e a r c h processes they replace. entire products around our pieces
division has had a mathematics of work.
department for more than 50 years. How predictable are the scenarios in
Our role is to work with clients on Cognos’ world of BI and performance What type of research have you focused
everything from supply chain and management? on in the past?
product design to demand forecast- We’re hoping to work that out with We’ve been involved in the on-
ing to [building] models to better the Cognos team—the acquisition demand strategy since it launched
understand and predict customer is [young]—and some customers to two years ago. We will accelerate
behavior. understand the problems that are that; we will have a lot more friends
in the IBM organization
How does that translate in that want to work with terms of the work you do on a ‘Our role is to work with us. We’ve been working
daily basis? clients on everything from with high-performance
We observe a business situ- supply chain and product scientific computing. We
ation—with supply chain I do work that supports
spent a lot of time in IBM design to demand forecasting.’ IBM supercomputers …
manufacturing, watching to make computers more
how it’s done. We decide usable. Information On
how decisions are made, then use at the cutting edge of what Cognos Demand is aimed at addressing
mathematics to figure out the best is able to address—[things such business problems.
decisions. We instantiate this in as] dashboarding or alerts, to put
software that is generally hooked more predictive modeling into and Where do you see your work with Infor-
up to ERP [enterprise resource to project [the] future. mation On Demand going?
planning] or CRM [customer rela- Computing has become much more
tionship management], and the Why is Cognos having such a big valuable as a [CONTINUED ON PAGE 45]