How do you bring together operational data from across the various divisions of a $148-billion-a-year company with 304,000 employees and hundreds of product lines? And how do you use the information to help customers discover new ways to use their products that have never been thought of before?
That was the challenge for Philip Kim, who was the marketing operations leader for GE's Measurement and Control division. I recently had the opportunity to catch up with Kim at the recent Tableau Customer Conference, where he discussed GE's burgeoning data analytics efforts. (Since the interview, Kim changed jobs and is now senior manager at Amazon.)
Even GE, which built a culture over the years based on innovation and analysis, is still in the process of discovering the potential of Big Data analytics. Kim led a team of analysts who oversaw the gathering of data to support GE's business units in marketing and helping customers identify new applications within its power plants, oil, energy and pipeline projects.
The challenge for Kim's team is not the volume of the data -- and there is a lot of it streaming in across GE's far-flung enterprise -- but its complexity, Kim says. "GE is a very large business, so we tend to deal with very complicated data sets." For example, he illustrates, the company may be selling power plant equipment to customers in Dubai and Brazil. "Just because they're both power plants doesn’t make them the same," he says.
To handle a world's worth of data and complexity, the key is to be able to deliver analytics fast and furious, working through many iterations and experiments, Kim says. The challenge for his analytics team is being able to turn around projects rapidly. "We take very complicated business problems, and then creatively come up with solution sets that are rapidly turned out. We're very fast," he says. "Or, as I tell my team, we're not necessarily smart, we're fast. I think its important to be fast to iterate through the various solution sets. And if we haven’t surprised ourselves or surprised our stakeholders, we probably haven’t done enough."
Speed to market is vital, he adds. "What we found very quickly early on was speed is essential. There’s really no reason why it should take a whole year. We focus on how fast can we fail, and as inexpensively as we can. We work in three-week sprints -- we try to figure out at the end of three weeks what's working, what's not. It means that we can't address world hunger, but we can figure out what's possible. It's the art of the possible."
Kim's team has performed analytic studies that not only look at GE equipment metrics, but also overlay it with information such as weather data.
Interestingly, while GE has long been data-driven in its decision-making, big data analytics is still relatively new, as it is for most organizations. "Within the last couple of years, we’ve seen a sea change. We’re starting to understand just how powerful what we are collecting we can use, when you look at the amount of data, and you combine that with the assume of a company like GE. We're starting to treat it as a discipline, much as we did with Six Sigma process control a few years ago."
Six Sigma -- which emphasizes lean methodologies and quality -- has been a religion at GE for a number of years, and this has also lain the groundwork for data analysis skills. "There’s no leader in GE than doesn’t know multivariate regression, ANOVA testing, cluster analysis, or two-tail tests," he says. "This is not common in most companies, and I’m talking about CFOs, CMOs and sales VPs. It's a pretty common language here."
Still, data analytics brings something new to the table at GE, he adds. "Analytics is now considered a new discipline for the new leader who's going to be in marketing sales. That helps them see beyond their skill set, to see other possibilities. Now they're expanding the problem scope, from a very straightforward repossess control problem into a broader business problem."
It also takes a special kind of person to be comfortable with this type of analytics, Kim says. "We try to hire people who are passionate about the work and passionate about data. It's just not people who know the mechanics of data mining. We want to find people who are very comfortable with an unasked question, and an ambiguous kind of answer. You have to find a person who’s okay with that."