Q: When Teradata was founded 30 years ago, a terabyte-size database was considered large. Today, we're starting to see petabyte-size databases. Just how large is a petabyte, and what does that say about the enormity of information available to manufacturers?
A: A petabyte is almost a number too big too imagine, but it's technically 1,024 times a terabyte, which is one quadrillion bytes of information. To put that into PC terminology, if every PC had a 50-gigabyte hard drive, then you'd need 20,000 PCs just to hold 1 petabyte. And we've got customers now that have 5 to 10 petabytes and growing. So the information explosion is here.
Q: You've said that the days of true operational business intelligence are here." What do you mean by that?
A: There's some new thought out there that the real competitive advantage going forward is going to be operational business intelligence (BI). By that I mean the ability to take your transactional and other corporate data that you need and be able to analyze that information and do it in near real-time and then to be able to drive the insight that you get from that operational BI down to the levels of decision makers. So it's building a culture of analytics and creating the enterprise IT infrastructure necessary to create the insight but then driving the insight down to people who can, on a daily basis, make decisions with facts and improve profits and grow revenue.
Tom Davenport's book Competing on Analytics: The New Science of Winning, pointed out that companies like Wal-Mart, Harrah's Gaming and Discover card have spent a lot of time and money creating [BI] capabilities and they now consider that competitive advantage.
Q: How are manufacturers operationalizing the business intelligence available to them?
A: At Ford, the business challenge was the sheer magnitude and the speed of their supply chain. Just a few years ago they were managing 250,000 parts, which would break down into 892,000 SKUs, and all of those parts were flowing through their system at different velocities, and everybody was just focusing on today's backorder problems rather than trying to prevent tomorrow's problems. So [Ford and Teradata] created the data warehouse and put the necessary detailed information into it, and an analytics team worked on predicting and projecting the inventory levels at any point in the supply chain by any part number. They created a predictive alert system in order to notify everybody of critical stockout situations before they occur. Then, operationally, they would recommend priorities down to the line workers so they were able to reduce their inventories, reduce stockouts and improve the overall efficiency of their supply chain.
Another good example is the large grocery store chains, which have become very efficient at maintaining and analyzing the point-of-sale data rolling off their scanners. Increasingly it's daily, but some of the big retailers are now taking that data on a 15-minute interval. And they are literally making real-time decisions and building analytic models looking for patterns of behavior, A in the supply chain and B in the demand chain, or in the consumer behavior. On the other side of that coin, you have manufacturers also interested in that point-of-sale data. Goodyear today is sweeping in point-of-sale from all of its retail outlets, and it's now using that to improve its supply forecasting. [Goodyear] has reduced its out-of-stock significantly, and it has streamlined its inventory positions, basically by using that same point-of-sale data just as the retailers do it.
Q: Is the enterprise resourcing planning (ERP) system the only data source for business intelligence?
A: Not in my experience. That's nothing critical about ERP systemsthe software today is more powerful than anyone could have ever imagined. But if you look at the kinds of information that manufacturers need, it's not just the ERP transactional data. For instance, in consumer goods, it's very likely that they are using syndicated data providers such as AC Nielsen and Information Resources to get demand information around their large retailers, and they use this to do improved forecasting, to look at effective pricing, to look at marketing mix, etc. Then they're taking that another step forward, going back to get the actual point-of-sale [data] from retailers, in order to better understand the demand patterns.
Another example is sophisticated high-tech manufacturing, where there is a great deal of information coming off the equipment, and [manufacturers] need to be able to do analysis on that to look at line efficiency. If you go industry by industry and look at the kinds of information that they're using to drive both their demand and their supply, it's not unusual to see some of it is not coming from the ERP systems.
Q: Why is it important for manufacturers to integrate data from a variety of sources, and why is it important to obtain information from additional sources outside the organization?
A: Well that's been a dream for some time. The promise of SAP early on was to build an integrated ERP, because customers realized 20 years ago that looking at information in a silo would only get you so far.
Teradata data warehousing takes this a step further and allows you to pull a myriad of transactional data and to analyze it in an integrated fashion and define things that you could only dream of previously. The demand for this continues to grow, because so many people in so many companies are pulling information from ERP and non-ERP systems and putting them into spreadsheets in order to answer business questions.
Teradata data warehousing allows you to do that on a large, systematic and standard basis in order to answer those questions.
Q: How can business analytics help manufacturers reduce the expenses incurred from product defects?
A: That is a cross-silo business problem. You have engineering, production, distribution, sales and then you have repair, and all of those functions can have some effect on where defects are going to occur and how quickly you can correct them. So in a perfect world, we would be manufacturing cars, they would be sold very quickly and we would have perfect information around where the defects were occurring. We would spot the defect quickly, we would go back to engineering, we would make a correction, we would correct the production line and then the cars going out very soon afterward would no longer have this defect, and we would minimize the payout for warranty claims significantly. For many major automobile manufacturers, that process can take anywhere from six to eight months or even worse, and then when you think about the number of automobiles being built everyday and sold, and the percentage that are going to have that defect, then you can see very quickly that there's opportunity for significant expense.
Teradata will allow you to create an early-warning system, so to speak, that can pull in all the information from the dealers and from the maintenance facilities and enable you to be able to apply statistical analytics against that data, spot warranty issues very quickly and then basically reduce that time cycle to get a fix to notify customers, to fix the production line and ultimately save significant money.
Editor's note: This is the first in a two-part series. Part 2 of IndustryWeek's conversation with Keith Henry will appear later this month.