SAS CEO offers analytics over BI, reveals use cases for text analytics

Organisations need to understand “big data” as a strategic asset, Jim Goodnight, CEO and founder of SAS Institute, told attendees at the software vendor's recent Premier Business Leadership Series event in Antwerp, Belgium.

Macy’s is one retail business that the Cary, N.C.-based analytics company has been helping optimise its sales operations, he said. Between Sunday night and Monday morning, millions of SKUs (stock-keeping units) are repriced to optimise revenue. “Imagine doing that by gut instinct,” Goodnight said.

Seeking to foster a reliance on data and analytics over gut feeling, SAS is offering a course to MBA professors to help them teach students to do more than use spreadsheets, Goodnight said. More generally, he deplored the shortage of young people in the West getting an education in STEM subjects (science, technology, engineering and mathematics). “If we want to remain innovative, we need to get youths into STEM. History graduates won’t create jobs.” He recalled that in the late '50s and early '60s, Western students were fired up by the space race to commit to mathematics. “We had Sputnik to get us focused.”

SearchDataManagement.co.UK interviewed Goodnight at the conference, and what follows is an edited version of that conversation.

What’s the value in making a sharp distinction between business intelligence and business analytics? Aren’t they on a continuum?

Jim Goodnight: I’ve always thought there was a difference. Business intelligence is the name that was given to simple query and reporting by companies like Cognos and Business Objects to make it seem more important and above what it really was. BI is query and reporting: You throw a query at your database, you get an answer, you print it out. Analytics is much more sophisticated. It’s about predictive modelling to help you forecast what customers might do or want in the future: what their behaviours might be if you introduce a new option for them and so on.

Analytics helps determine whether drugs are effective or not. There is nothing in BI that will enable you to get a drug through the FDA approval process. You’ve got to prove that the drug is effective; that it improves treatment over the currently used drug. You can’t drill down in a BI report to find out the probability that a customer will leave in the next month. That’s churn analysis and telcos do a lot of that. You can’t drill down in a BI report to find out the probability that a customer is going to default on their loan.

BI just gives you certain numbers that report on what’s happened in the past. Sometimes they throw in graphs so you can see what’s happened in the past. But if you are in charge of deciding to accept the charge that’s just been made when someone has swiped their card through a reader, that’s very sophisticated. Is it fraud or not? That’s what we do. There are very sophisticated models behind that, based on neural networks. Now, that is analytics.

There was a theme at the conference this morning that the time of analytics has come again. There are those books by Thomas Davenport and his colleagues, published by Harvard, on analytics that have been in vogue. Why now for analytics?

Goodnight: I think more and more businesses have adopted analytics as part of their everyday processes. When SAS was first founded in 1976, our main users were university researchers doing experiments, pharmaceutical companies and a few insurance companies doing actuarial work. Other than that we didn’t have a lot. But over the years others have used analytics. Banking was the first industry to really get into predictive modelling. And today most banks have several thousand models that are actively telling tellers what to do and helping them make decisions: deciding what the current formula is for computing risk on loans, helping establish the right rate of interest for a person.

So it’s been an organic expansion from that core?

Goodnight: Yes, one sector that is interesting is retail: using point-of-sale data and loyalty card data to come up with much better understanding of customers. Size optimization is also a big area in clothing. We look at a year’s worth of data [for a retail company] and help them decide on what sizes they should carry for an individual store. You don’t want to order the wrong package, else you have clothes left over or not enough. So a simple thing like that can make the store and customer happy.

I know that the underutilization of SAS technologies is something that puzzles and concerns you. What are your thoughts on this relative lack of exploitation?

Goodnight: A lot of times people are just now aware of just how much power is in the SAS technology that they already have. I’d like to see them use more of it. We have some places that have 2,000 users and others that have only a few.

I’m also still involved in R&D at a pretty detailed level. I’d like to get more involved in our in-memory analytics, where we spread data over a thousand different machines and work on it in parallel. And I want our own BI people to use the same technique. I want a completely in-memory server for BI. We already had one that our BI people were not aware of, and I had to explain it to them. The in-memory development was done in our advanced computing lab, you see.

What’s the place and value of text analytics for SAS?

Goodnight: One thing we looked at was bank call centre records around foreclosure. Who do you keep? Whose loan do you turn over to a collection agency? There are six or so categories that people fall into. Banks have models to put people into the right category so that they know which approach to use. We found that we could improve the performance of those models by about 50% by using text analytics on the calls.

But the big thing we are doing with text analytic lies in social media analysis: looking at blogs and tweets about brand names to determine whether positive, negative or neutral. And there is a proactive aspect to that, in our Twitter chat room, where people contact those saying bad things and try to win them over offline.

Warranty claim analysis is another area. This is where a company carries out a repair under warranty and bills the manufacturer. We are using text analytics on these claims to add variables that help determine whether they are fraudulent. We take all the claims made fraudulently by hand, collect the data and build a model to detect fraud instantly.

The theme of the conference is ‘Innovate. Optimize. Transform.’ You gave an interesting example of optimization this morning with Macy’s. Do you have a good example of what you are doing that is innovative?

Goodnight: We are working with oil and gas companies on predictive asset maintenance. We look at all past failures of parts on the oil rigs in the North Sea and use that information to predict when parts will fail. We can give an accurate forecast on when a part will fail. So if there is a part that hasn’t failed yet, but there is some downtime coming up, then they will schedule to replace that part.