Island Data, a California firm founded in 1995 to help call centers automate some functions, revamped its business in 2004 to create software for scanning e-mails and other customer feedback for trouble, attrition risks and up-sell opportunities. The firm, whose clients include Wal- Mart, Microsoft, Yahoo and Best Buy, has raised $6.5 million from venture capitalists and angel investors. Its real-time scanning of customer e-mails is trickier than it might seem. It requires algorithms to create a self-building system for analyzing the text in a message. “This is still an early stage market, but it doesn’t take a huge leap of faith to see that listening to customers better creates an advantage,” says Guy Jones, Island Data’s president.
Island Data is one of several companies using sophisticated algorithms to analyze data to make better business decisions, a field known as analytics. Analytics is expected to grow as more companies figure out how to better use to their advantage the mountains of data they have collected over the years.
Building on research on neural networks done at the University of California San Diego, professor Robert Hecht Nielson co-founded HNC Software to develop software to uncover credit-card fraud. It recognized patterns in the behavior of each credit-card user, helping it to sniff out when something might be out of whack. HNC’s software became the dominant product for screening credit and debit cards. The firm was purchased in 2002 by credit- scoring giant Fair Isaac for $810 million, but over the years several executives and scientists have left to start their own firms, applying pattern-matching algorithms to fraud problems for mortgages, cell-phone transactions and a host of other related issues.
“We’re strongest here in San Diego on the more advanced analytics, where you’re using more mathematics and statistical analysis of the data, rather than just the creation of reports,” says Michael Chiappetta, a consultant who formerly worked at HNC/Fair Isaac. “I think predictive analytics right now is one of the more advanced areas, where you have a statistical probability of some future event occurring, whether that’s a statistical probability that fraud is taking place or a statistical prediction that somebody will respond to a certain marketing offer,” he says.
At the University of California San Diego’s Jacobs School of Engineering, professor Gert Lanckreit is leading an effort to create a “natural language” search engine for music that would allow people to find music they like without knowing the name of the artist or song. For example, a user could search “upbeat music with female vocals” and the search engine would return a list of songs. The hard part for computer scientists is labeling millions of songs accurately, an almost impossible task to accomplish manually. But over the past two years using analytics, Lanckreit and his team have developed a series of algorithms that have taught computers to pinpoint features in songs. They call it a computer audition system. Once features in a song are annotated, the search engine can retrieve the song. The research is an example of how analytics can be used beyond fighting fraud.
Scott Gnau, chief development officer for Teradata, a data warehouse and data-mining software provider, says many businesses have gathered reams of data. They have the back-shop technology to access it and the software to generate reports. More of them are now using analytics to tap this data to make real-time business decisions every day.
For example, a financial institution might see an unusually large transaction and use automated software to screen it for fraud or money laundering. But after that, the bank might also want to take advantage of the event to create customer loyalty, Gnau says.
“To do that, you need to understand the customer—their history, all the different touch points of the customer and the demographics of the customer—on a very specific level. The leaders in the marketplace are doing this today,” he says.