Prescriptive analytics evaluates decision options in uncertain conditions and their expected outcomes and impact. With big data, it crunches data comprehensively, at varying levels of granularity, for decision-makers to bring light to gray areas in decision-making. By reducing latencies in the assessment of options, executives are able to continually reexamine assumptions. They gain a multidimensional situational awareness that minimizes the prospect of executives being blindsided by a narrow view of their choices. Prescriptive analytics rapidly parses data to answer “what-if” questions to anticipate challenges, make course corrections, and preemptively respond to worst-case scenarios.
Prescriptive analysis is the pinnacle in the maturity curve of analytics, and 2016 is expected to be the breakout year for it following the success of predictive analytics. Currently, only 3 percent of users of analytics take advantage of prescriptive analytics, compared to 30 percent who use predictive analytics. “Predictive analytics is more narrowly focused compared to prescriptive analytics. While prescriptive analytics often uses the same raw data as predictive analytics does, but more likely the forecasted data from predictive models, the volume of raw data expands as a gamut of physical data such as bill of materials is also included,” Eric Kelso, vice president of Product Management at River Logic explained to me.
Corporate prescriptive analytics Corporate strategic and tactical decision making, investments, mergers and acquisitions, and capacity planning all require decisions that can be optimized through the use of prescriptive analytics. All too often, individual departments like sales make deals that could have downsides for another department like finance. Similarly, companies can overlook the opportunity costs of alternatives foregone. An integrated view of decisions ensures that the company overall benefits from prescriptive analytics driven optimization.
My conversation with Eric Kelso revealed a striking business case of business gains realized as a result of prescriptive analytics that uncovered unsuspected opportunities. Cox Industries, a River Logic customer, increased its margin by over 2% after the wood manufacturing company decided to abandon a long-time business unit as its largest customer became more demanding. “What-if analysis revealed that new business opportunities would be much more profitable than keeping that existing business,” Kelso said.
The IoT and real-time prescriptive analytics The Internet of Things expands the scope for “what-if” analysis with microscopic data for optimization of operations. According to ABI research, the share of revenues from prescriptive analytics with M2M data is expected to rise from 1 percent in 2013 to 9 percent in 2018. Data from the Internet of Things is especially useful for real-time decision making in crisis situations or those prone to be so.
Unsurprisingly, the early uses of prescriptive analytics are in trauma care in the healthcare industry, oil and gas industry, and the supply chain. The Alfred Hospital in Australia developed a real-time response system to cope with the critical first 30 minutes in trauma care. In this short period, a team of physicians, nurses, and other staff have to work as a team, gain situational awareness and attend to injuries before irreparable harm is done. The hospital’s algorithm is the key to rapid and effective response; it is trained with evidence from past cases and medical research. Its real-time data input is vital signs data of the patient as well symptoms and the data on the injury. Based on the real-time data, the algorithm churns out recommendations for therapies. The medical team is readied to respond quickly for surgery with visual display of the information and possible therapies.
Consider the potential for prescriptive analytics in oil and gas. Shale oil is characterized by wide variance in productivity of individual oil wells. Drilling is more productive with precision targeting, but that takes more time to plan than boring over a wider area. Precision targeting also tries to find a focal point with clusters of oil surrounding the well bore that will maximize yield. A related consideration is the spacing between the well bores; they are more likely to have an adverse effect on each other if they are too close. The real-time data from acoustic sensors is used to make the assessment of alternative courses of action. It is combined with large volumes of geological and geophysical data, 3D imaging, and structured data of wells to gain insights. The data on the progress of drilling generated in real-time assists in reviewing the initial assumptions and postulating new hypothesis for increasing productivity.
Fitting into the supply chain
Supply chain is the natural sweet spot for real-time optimization with a gamut of operations decisions for capacity utilization, scheduling of cargo deliveries, and demand fluctuations. I spoke to Jim Hayden, Vice-President, Solutions, at sensor company Savi to learn about decisions and trade-offs addressed by prescriptive analytics with real-time and historical data for supply chains. “Our data scientists train our algorithms with historical data and our software triggers them with real-time data feeds of events from the Internet of Things,” Hayden said.
The algorithms assist in providing a variety of services customized for each stakeholder in supply chains. “For manufacturers, Savi’s algorithms enable the optimization of sources of materials and raw materials by estimating the expected time of arrival for each of them,” Hayden said.
By contrast, the retailers want to weigh the trade-offs of timely delivery versus the transportation costs that rise inversely with time allowed to load cargo on trucks. “Real-time data crunching of data helps to match the flow of demand at retail stores with the supply as it is affected by the pace of deliveries,” Hayden explained.
Finally, the multimodal shipping companies want to control the demurrage costs incurred at the ports for the storage of cargo before it is moved out by trucks. They can minimize these extra costs that are tied to the use of a ship while it is in port, by synchronizing inbound shipments at the port with the outbound haulage by trucks. The application uses marine traffic, shipping, sensor and satellite data to estimate the expected time of arrival for ships and the estimated time delays at customs, "depending on the type of cargo to synchronize the availability of trucks to shorten the time spent at the port,” Hayden said.
Algorithm driven prescriptive analytics lowers the latencies in decision-making caused by assimilation of complex and large volumes of data. It reduces the data to simple metrics to evaluate trade-offs and visualizes the impact for rapid assimilation. The early use cases can only whet the appetite for more.