Steven Mills, senior vice president in charge of software and systems at IBM, has seen just about everything since joining the company in 1973: Mainframes, PCs, OS/2, Lotus, analytics and Watson.
We caught up with Mills at IBM’s Madison Avenue office. What Mills brings to a conversation is historical context. A lot of so-called new ideas are reformulated versions of older computer science developments.
Here’s a recap of my chat with Mills.
On analytics and the hubbub surrounding it: Mills said that analytics for businesses is part of a “perpetual hunt to do better.” He noted that the hunt is on for new patterns and algorithms to digest the influx of data. I asked whether analytics was a work in progress. Jim Baum, head of IBM’s Netezza unit, had said that analytics work has been focused on infrastructure, but the front-end user interface will become crucial.
It appeared that Mills didn’t quite agree with that take, but didn’t say so outright. “Report generating does a fine job today, but in many cases that’s not new analytics,” said Mills. “The problem is to get data organized, the right data structures and consistency.”
In other words, analytics can only go so far without data management. Mills referred to a chat by Jeff Jonas, an IBM data wonk. Jonas argues that accuracy improves with data quantity. Analytics helps you sort through it all. “More data means more accuracy. Statistics 101,” said Mills.
Mills scoffed a bit at the argument that analytics was an underserved market. Hewlett-Packard CEO Leo Apotheker made that case at its strategy day. After citing customers like USAA Insurance, which was using analytics to better interact with customers, Mills said the promise is that new tools can make the customer relationship two way and “more event and behavior driven.”
Is business intelligence different than analytics? I asked this question because if you’ve been around this industry long enough you quickly find out that old technologies get respun into something exciting and new. Cloud computing is the offspring of the mainframe. Mills' response was worth noting.
One person’s reporting tool is another’s analytics. We don’t sweat the terms, but it’s all the same. Analytics has moved to be more real time and today the bulk of what people do is retrospective. The goal is to migrate from retrospective to figure out what will happen in the future.
Where does Watson fit in? For IBM, Watson highlights the future of analytics. Mills said the aim is to take large quantities of information with ambiguous characteristics and find meaning. As far as applying Watson to markets, healthcare is an obvious example. In fact, healthcare was “the perfect example” of how Watson in its current form can be used, said Mills. In healthcare, a patient defines sometimes vague symptoms and a doctor has to consult data, experience and literature to figure out the cause. Tests are ordered. It’s complex. Watson’s aim is to “use the diagnosis pathway to improve delivery,” said Mills. “If Watson could assemble and bring all the data together there would be more accuracy,” said Mills. “What healthcare professional can read all that data for a quick diagnosis?”
Physicians will need Watson as an assistant to scour data from web sites, statistics, and information from the source. “As physicians feed it information it iterates and can lead to outlier cases,” said Mills. “Watson can hone in on qualifying factors and through iteration get better and better.”
As for other verticals, Mills noted Watson would have to be tweaked a bit for “simplified versions that can run any sophisticated call center.” In this scenario, a baby Watson would be used for institutional knowledge capture in call centers and engineering outfits like Boeing.
The role of in-memory databases. In-memory databases, which promise to crunch data faster, “are nothing new,” said Mills, who seemed a bit amused that in-memory applications were generating so much buzz. IBM was tinkering with in-memory applications in the 1960s. What changed today is that the economics make more sense since memory and storage are closer in price. Mills also mocked HP (HPQ) for buying Vertica and probably paying too much (terms weren’t disclosed). “In memory is so mundane. If it makes sense we run an application in memory. Duh,” said Mills. He said that in-memory will be used for workloads that need speed, but other forms of databases will still be around. Data will be tiered based on whether it’s archival or information that’s needed in real time. Companies will optimize a mix of in-memory, solid-state storage and disk, he said.
“In memory will be used in combination with other storage,” said Mills. The buzz around in memory can lead to thinking that it’s good for everything. “You can’t decide that in memory is the way to do it and then find out it’s only for a subset of work and a bad idea for the rest of workloads,” said Mills. “In some ways you can get caught up in the adverse effects of a good idea.”
IBM’s acquisition strategy. IBM has been extremely acquisitive and gobbles up software companies at a rapid clip. Mills, however, doesn’t go for the big expensive bang. “For many years, IBM has been building out an architectural approach to solving customer problems,” said Mills. “We are not an accumulator of companies. This is not a jewelry ensemble.” In other words, acquisitions have to fit in with an existing business or fill in gaps. “Customers are good at voicing gaps and that leads us to make decisions about where to make, buy or partner,” said Mills. “We make a lot of stuff, but do partnering than buying.” IBM’s acquisition parameters go like this:
- Does the acquisition fit?
- Can it be accretive in two years?
- Does it fill a need? “We’re selective about what we buy and many are small companies pre-IPO,” said Mills.
IBM’s recent example of Tririga, which makes real estate and facilities management software, fits the acquisition mold. Tririga fits with another previous acquisition, MRO, which tracks assets. Mills likes to buy companies that the average bear hasn’t heard of.
Would IBM buy Lawson, which is a target of Infor? “We’ll see what happens with Lawson. We do work with Infor and Lawson so we assume we’ll work with whatever comes out of that,” said Mills.
What acquisitions are off limits? Mills said there are areas where IBM won’t go shopping. “We’re sensitive to ecosystem implications. You can buy a company and end up being a minority player where the majority hates you,” said Mills. IBM thinks about the business it might lose in other units such as hardware and services. In short, IBM has no interest in buying a company like SAP. “Services does a huge amount of business with SAP and Oracle,” said Mills. “You can’t compete everywhere.”
What are CIO concerns beyond analytics and cloud computing. Mills listed the following:
- Business process integration. “Customers spend more money on business process integration than anywhere else,” said Mills. “How do you bring applications together to support the business?”
- Application development and lifecycle management.
Social enterprise software. Mills said social software has the promise to do what knowledge management systems couldn’t—capture user knowledge. “Inside the enterprise, social networking lives in human time as opposed to process time,” said Mills. “Social networking provides a mechanism that’s adaptive to the way people want to work in style and approach.” He added that enterprises are open to social networking because it provides an outlet for finding experts, community driven help and a historical record of relationships. The rub: These networks have to be secure. “What works on the public Internet doesn’t work in business and government,” he said.