Main image of article The Rise and Fall of Black Box Analytics
As businesses of all sizes come to appreciate the potential of advanced analytics applications, there’s a rising temptation to use these technologies to gain a sustainable competitive advantage over rivals. But analytics applications also offer a significant challenge: the analytics powering businesses practices often do their thing within a proprietary “black box,” hidden from the client’s view—raising the specter of a scenario in which an algorithm, improperly vetted, wreaks havoc. Indeed, behind most analytics applications stands a small team of engineers and developers with a high degree of specialized knowledge. The resulting algorithms are pretty arcane, meaning that not many people really understand how they work. Nonetheless, such understanding is necessary: there’s more pressure than ever, at least from a compliance perspective, for senior managers to clearly document processes. Those opposing forces have created pressure on IT vendors to build analytic applications that are not only more transparent, but incorporate the expertise of a much broader number of individuals. “Collaboration and compliance are both going to become major issues in the coming year,” said Tom Cowan, CEO of Decysion, a provider of analytics applications headquartered in Stamford, Conn. that recently picked up $15 million in funding. “The ‘Black Box’ era of analytics is dead,” added George Mathew, president of Alteryx, a rival provider of analytics applications based in Irvine, Calif. “One reason we’ll see so many analytics applications running in the cloud is that it provides a way to more easily collaborate on the development of these applications.” In fact, rather than relying on hard coded models, Deepak Advani, IBM vice president of business analytics and solutions, suggests the need for collaboration and compliance will become major drivers of the future, one in which organizations will set up an internal “analytics factory.” Not only will such an approach bring more transparency to how analytics algorithms are developed and deployed, Advani added, but an analytics factory will make the whole process more economically efficient: “Right now analytic applications rely on small teams of people that take a long time to develop an application that is usually hard coded, which means the applications usually has limited value from a reuse perspective.” IBM envisions a world where teams of people with varying levels of expertise come together to develop not only applications that result in a better particular business outcome, but also intellectual property that can be reused across multiple applications. But achieving that goal by definition requires a lot more collaboration and transparency, Advani said.

Predictive Analytics

Much of the need for greater transparency is being driven by increased regulatory requirements. Don Busiek, general manager of manufacturing software at GE Intelligent Platforms, noted that manufacturers now must provide “genealogies” for the products they create, including the ingredients used in the manufacturing process. While that might seem onerous, he noted that many GE customers are starting to run that information through analytics applications, the better to identify potential risks to business: “The future of advanced manufacturing is going to be all about leveraging analytics.” According to Adam Binne, global vice president and general manager for SAP Business Intelligence, the shift towards prescriptive analytics is still in its relative infancy. Most organizations are squarely focused on descriptive analytics that generate more accurate information. A few others have embraced proactive analytics that take advantage of suggestions and recommendation engines to make better decisions. Only a handful of platforms have achieved a prescriptive analytics capability that allows them to automate an end-to-end business process. “The biggest issue most organizations are still wrestling with is simply acquiring the analytics skills needed,” said Binne. Nevertheless, interest in prescriptive analytics capabilities is growing. Business leaders across the spectrum have seen how financial services firms have leveraged analytics algorithms to trigger a whole series of related events in a matter of sub-seconds.  Despite the risks, many business leaders want to apply similar capabilities to automate process in real time, which (at least in theory) could increase profits by orders of magnitude. One of the things that make it more affordable for organizations of any size to apply these concepts more broadly is the advent of more affordable means of processing Big Data. Instead of merely sampling data, suggests Dr. Satya Ramaswamy, vice president and global head for TCS Mobile at Tata Consultancy Services, organizations can now make decisions based on all the raw data they have at their disposal. “It’s now possible to exhaustively go through all your data to come up with a definitive answer,” said Ramaswamy. Beyond building those applications, the next major challenge involves getting the results of those queries out to end users’ mobile devices. In short, the coming year will witness the emergence of a new class of collaborative analytics applications that are not only more transparent from a governance perspective; they will be more strategic in terms of the number of business processes they actually help automate. But while there may be a lot of business pressure to reach this new era, organizations might do well to heed the advice of Warren Buffet when evaluating complex IT applications: “Beware of geeks bearing formulas.” The best way to do that, of course, is to make sure that every advanced analytics application is vetted by people capable of plunging into the deep weeds of algorithms and business processes.   Image: gonin/