Using statistical analysis of data points collected in real time during the process, Amgen identifies "weak signals" that could indicate brewing problems in the manufacturing cycle. Scientists then delve deeper, taking corrective steps if necessary. The analytics system includes virtualized data warehousing tools from Denodo, multivariate analysis tools from Umetrics and various software modules from SAP.
By modeling multiple potential variations that could be indicated by the weak signals, Amgen can draw conclusions much faster than in the past, says CIO Diana McKenzie.
In one recent example, Amgen identified the cause of a cell culture problem about a month earlier than it otherwise might have. For that biologic product, making the fix early--and not losing that month--saved $2.4 million, she says. Depending on the product, manufacturing a lot can cost more than $1 million, she says. "This is Amgen's taming of big data."
Steve Randich, CIO of the Financial Industry Regulatory Authority (FINRA), is slaying a data analytics monster of his own. The primary mission of FINRA is to protect investors from Wall Street crime. But investigating alleged Wall Street criminals, even with politics aside, is immensely complex.
FINRA keeps a few dozen structured and unstructured databases and other electronic collections of information on about 4,250 brokerages and 630,000 brokers and other registered financial advisers. Employees can search all these sources for material related to an individual or company under review, developing multi-stage queries that weight data sources depending on the kind of information they're after. They can also set up alerts so they are notified when new pertinent data is available. But it didn't used to be that way.
Until 2011, some FINRA employees didn't have immediate access to some sources. Other critical files were still paper, stored in boxes and cabinets. It could take weeks to complete a basic background search. FINRA decided to digitize most data sources and build its own search capabilities. The Enterprise Search engine, built on Solr open-source software, does full-text searching like Google, but is tailored to FINRA's need for "intelligent" filtering and weighting of information, Randich says.
Now a typical search takes less than a minute, allowing more time for more investigations. FINRA suspended or barred 843 individuals last year, compared to 684 in 2008, a year when it logged 94 percent more complaints.
"If you're getting information faster and easier, the cost of doing so is going to be less," he says. "Investigating a case is easier, cheaper, quicker and more likely to result in catching the bad guy."
Focus Beyond IT