In 2012, for example, Intel IT created a new reseller sales tool that worked to increase the chip maker's revenue by enabling its sales team to identify, then strategically focus on, larger-volume resellers. The new software engine mines large sets of internal and external data, then applies a predictive algorithm to pinpoint the most promising resellers. So far, it has helped identify three times as many high-potential resellers in the Asia-Pacific region as manual methods typically would have uncovered, according to Stevenson. That translates to about $20 million in potential new and incremental sales. More gains are expected as the tools are rolled out to other geographies.
On the manufacturing front, Intel is using a predictive analytics tool to reduce microprocessor testing time. The company saved about $3 million in testing during a proof-of-concept period. By 2014, as the tool is implemented more widely, Stevenson expects it to rack up another $30 million in savings companywide.
Intel's analytics success has been fast-tracked, to say the least. The key, Stevenson says, is tackling big-money problems with relatively small and swift-acting teams.
"To get the business to focus on the future and ask better questions that would lead to better outcomes, we knew we would have to do things quickly," she explains. "We were coming out of a traditional BI environment where solving master data is the unsolvable problem. People work on it forever and the business doesn't necessarily see the value."
So Stevenson came up with the "six months and $10 million" rule. "A $10 million problem solved in six months is important. Any general manager would say they'd invest six months if we could save them $10 million," she says. (At Intel, business managers must support and fund IT projects.)
Stevenson recruited five-person teams made up of a business expert, a statistician, a predictive modeler, a machine learning expert and a data scientist. "Each person on the team had a slightly different perspective on the problem we were trying to solve. Doing it in six months was our way of earning the right to prove the capability was there to really change the way we do things," she says.
In addition to the projects that reduced testing time and pinpointed lucrative resellers, 13 other analytics projects have been completed using that approach. So Stevenson has upped the ante by finding $100 million problems and challenging teams to solve them.
"When you have a track record, you can ratchet up," she says. Other ongoing projects include a predictive engine for streamlining Intel's chip design and debugging process and another to predict new information security threats.
But Stevenson cautions enterprises not to underestimate the skills required for analytics initiatives and the time it may take to nurture those skills.
"When I think about our learning curve with Hadoop and some of the more advanced presentation layers that are very different from SAP or traditional BI, I'd emphasize that there is a learning curve there for technical skills that isn't insignificant," she warns.
Her other piece of advice: "Develop an appetite for experimentation," especially since analytics technology is still evolving. "The winners and losers on the tech side are not completely shaken out yet," she says. "Keep your aperture wide."
Read more about applications in Computerworld's Applications Topic Center.