In 2011, a small team within Intel’s IT department started looking for ways to more accurately predict the computing demand required to test chip designs.
By knowing when certain resource-intensive steps would take place and how much memory each test would require, Intel could make more efficient use of its large computing server grid, scheduling more concurrent tests without running out of memory or processing power.
The stakes were high: For every 1 percent of efficiency gained, Intel would save $1 million a year. The team of five spent six months applying advanced analytic techniques to data from test simulations, finding design process improvements worth more than $10 million a year.
The team, now known as the Advanced Analytics group, has been swamped with project requests ever since—even as the group has grown to approximately 100 members. To prioritize proposed projects, it analyzes the strength of business unit support and collaboration, the quantity of high-quality data, and the size and extent of the potential impact, among other factors.
Today, projects the team deems worth pursuing could last a year and yield returns much greater than those of the chip design simulation project.
Demanding high returns on projects “makes very clear that this is the order of magnitude of value that we are expecting,” says Moty Fania, Intel IT’s principal engineer for big-data analytics.
The team’s work still includes plenty of exploration and experimentation, though. “Almost always you start looking at the data with a hypothesis or assumption in mind,” he says. “But sometimes you find a different signal in the data that you weren’t expecting, and that becomes the important thing.”
But sometimes you find a different signal in the data that you weren’t expecting, and that becomes the important thing.
2. The right problem. The question the analytics team is seeking to answer must matter to the business. Addressing this problem also requires understanding how any solution fits into the organization’s existing business processes, systems, and staffing arrangements.
In this case, the problem aligned perfectly with the organization’s mandate to increase sales. Top managers pledged to carry out business process changes, if necessary.
3. Data. The analytics team has to determine if there’s enough high-quality data to make an analytics project feasible and worthwhile.
The team carried out a pilot project for one geographical area of the inside-sales organization that managers knew had a high-quality data set. Successful tests with this data showed that the analytics project held promise, which justified work to upgrade the data quality in other areas.
4. Resources. The analytics team, working with the sponsor organization, has to evaluate what individuals, skills, tools, and processing power are needed for the project to succeed, and whether they’re available.
The analytics group had a team in place working with the inside-sales function. And sales leaders made people available to the analytics team.
5. Time. The analytics team should assess whether it can achieve results for the project within a desired time frame. You want to quickly demonstrate value to the organization.
For the inside-sales example, a pilot project yielded strong results in short order. This justified more work on a broader data set.
In many cases, the conversation starts in one place, and we end up tweaking it, and we end up with a different solution.