Gather IT executives from different industries to talk about the challenges of harnessing value from big data and you will hear common themes: The size of datasets to manage. The complexity of managing projects to produce results. The vendor relationships to finesse.
But listen further and you will hear tips for meeting these challenges. At an October 22 panel discussion at the Big Data Conference in Chicago, executives from the insurance, technology and health care fields talked in straightforward tones about aligning big data analytics goals with corporate objectives, their use of pilot projects, the need to narrow datasets to speed progress, and the emphasis on vendors having skin in the game.
“We’ve mapped the [technology] value proposition to the business value chain,” said Rachel Higham, chief enterprise architect at ACE Group. Established in 1985, ACE Group has evolved into one of the world’s leading providers of commercial property and casualty insurance with nearly $24 billion of gross written premiums.
Higham, who also manages the innovation, research and development activities for ACE Group, emphasized the importance of conducting proof-of-concept pilots.
Typically, these pilots take a subset of the company’s largest datasets, which can be billions of rows for its commercial customers and trillions of rows for retail customers. Moreover, they are always explained to management in terms of their business benefits, Higham said.
In fact, the insurer has established an “Innovation Fund” to fund these proof-of-concept projects, which are showcased in high-visibility events for both the business and IT teams.
In one example, Higham’s group demonstrated how IBM’s Content Analytics could improve the company’s understanding of known factors driving high claim amounts, something other ACE analyst teams might spend several months working on.
“With only 80 hours using the IBM content analytics, we were able to match those results for smoking and diabetes,” Higham said. The analysis also found claims correlation with other factors, such as obesity, sleep disorders and substance abuse.
“In addition, predictive analytics with ICA discovered previously unknown factors and highly correlated with claims in excess of $100,000 related to disease symptoms and medications,” she said.
Moving Ahead with Less Data
Virendra Vase, CTO at social media-influence-ranking company Klout, likewise touched on the practicalities of big data projects.
The company, which has scored more than 400 million people in terms of their social media “influence,” now deals with 5 petabytes of data, and processes 12 billion “signals” daily, Vase said. “We’ve built core capabilities in data and science that cannot be replicated,” he said.
Nevertheless, Vase said any commercial entity must appreciate not only its end goal and success metrics, but the tradeoffs that are necessary when time-to-market is a strategic priority. For example, a company managing and analyzing large datasets must assess its priorities with data-centric needs in mind, such as the size of its datasets, the query performance it demands, the query patterns it expects to experience.
Appreciating those trade-offs can help determine which technologies to use, Vase said. And like ACE Group’s Higham, Vase said he recommended a culture of fast iteration with smaller data sets.
Co-Creating on CRM
Data analysis drives and verifies the efficacy of marketing efforts at Tenet Healthcare, the nation’s second largest healthcare network, said Brian Barnes, vice president of consumer applications. But making this work has involved a lot of vendor management.
“It took eight companies to this together,” said Barnes said, adding, “Best-in-breed can mean a full barn.”
Barnes said the resulting marketing platform, which includes internal and some purchased demographic data running on a Hadoop cluster, has resulted in a revenue lift of nearly 1 percent.
While Tenet Healthcare, which has a network of 80 hospitals and 157 outpatient centers, has fully outsourced its big data effort with its CRM vendor, it found a way to keep this vendor close and engaged. It jointly created with this vendor a new company to handle the work. Barnes did not name the vendor, but he said Tenet has an investment in the joint venture.
Weather Data Apps
Many speakers at the three-day big data conference, which was co-located with an event called Cloud Connect, discussed how analytics, combined with cloud services and mobile delivery, was ushering in a new generation of commercial products.
For example, Paul Walsh, vice president of weather analytics at The Weather Channel, revealed during a keynote that his company is quietly rolling out a pilot service that will compute local weather for mobile users “on the fly,” based on the nearest weather station. General availability of the as-yet-unnamed service is set for 2014, he said.
Such real-time, targeted prediction will open up new business opportunities, from farming to public safety, and will change the traditional paradigm of weather prediction from “cope and avoid” to “anticipate and exploit,” Walsh said.
Indeed, people living in Pakistan and Bangladesh are already better protected during monsoon season because of advanced warnings delivered to their cell phones. A flood 15 years ago in the region killed 15,000, but during a similar cyclone earlier this year, “very few were killed,” Walsh noted.
Ellis Booker is a freelance journalist based in Evanston, Ill. Email him at firstname.lastname@example.org. Follow him on Twitter: @EllisBooker.
Home page image of mountain climbing steps by Matt Wakeman via Flickr. Used under Creative Commons license.