Traditional Business Intelligence (BI) is a rear-view mirror that helps companies understand decisions and customer actions that occurred in the past. Predictive Analytics looks ahead, allowing companies to make the timeliest and most effective decisions today.
Predictive Analytics is a complicated process that can bring huge payoffs, but which also has enormous implications for the IT infrastructure, business decision-making and how people interact in your organization.
In an age of social media and the digital economy, information can travel around the world in seconds, and customer desires can change on a dime. In this fast-paced environment, it’s critical for companies to be able to anticipate needs, not just react to them.
For example, the healthcare industry is under extreme pressure to reduce costs and improve efficiencies. It’s simply not enough to make sick people better; the goal is to prevent people from getting sick in the first place. One key area that hospitals are seeking to improve is readmissions, where patients return to a hospital within 30 days often because their initial problem wasn’t solved.
With this in mind, a large hospital used predictive analytics to pinpoint patients who had a high readmission risk by leveraging a wealth of information, including their electronic medical records and socioeconomic data. As a result, the hospital realized a reduction of readmission rates, saving in medical costs and avoiding potential Medicare penalties. In addition, this freed up resources to serve many more patients1.
The power of predictive analytics is that it helps companies target their efforts, so they can get the maximum results with the least amount of resources.
Using analytics to reliably predict future behavior can do more than allow a company to increase sales or improve efficiencies. The unprecedented level of actionable insight from predictive analytics can also enable companies to re-imagine their business models and basic value proposition. Used adroitly, predictive analytics can erase barriers between industries, opening up entirely new profit streams.
In Italy, for example, insurance companies face the highest frequency of auto accidents and highest average claims in Europe. At the same time, price-comparison websites have intensified competition, tightening profit margins on policies.
In response, one large insurance firm installed black boxes in their customers’ cars, which gathered, stored, and analyzed data about their driving habits, such as how fast they took corners or how smoothly they applied brakes. This allowed the insurer to predict their likelihood of having a crash, and also let the insurer incentivize poor drivers to change their habits.
While this approach fine-tuned the insurer’s existing business, predictive analytics has helped transform their business. As a result of the actionable data, the insurer has developed new business models, called “Pay as You Drive” or “Pay How Your Drive,” which charge people based on their actual driving habits rather than on more traditional and more hazy predictors like age, gender and driving experience. At the same time, the information allows the insurer to roll out new services, such as fuel management and remote diagnostics that previously were outside the scope of their industry.
When developing the business case for predictive analytics, think beyond improving your current processes. The power of predictive analytics can be in providing new insights that transform what you do—moving you into profitable new areas that may disrupt your industry.
A fundamental theme of predictive analytics is agility – the ability to anticipate events and opportunities. However, successfully implementing these projects requires a strong alignment of IT and Business Units, with business decision makers involved in the R&D phase of an analytics initiative.
Consider EMC*, a leader in storage and analytics technology, which found itself unable to do target marketing and lead generation effectively because its own data was siloed in too many places. EMC* had acquired 80 companies over a 10-year period, leaving data scattered in many forms, in many locations. The company consolidated many siloed islands of data into a data lake, from the unstructured data from social media to structured data like customer records.
A key part of the process was not simply focusing on the technical elements and infrastructure requirements for predictive analytics, but considering how to support each business unit’s need to understand the data. For starters, EMC* implemented governance that allows business groups to share and collaborate on analytics projects. Each line of business was also given a “sandbox,” where it could run analytics scenarios in a self-service environment.
This shifted IT from being a gatekeeper of the data to a facilitator of insights. IT provided technical and data expertise to business units, while freeing the business units to take the lead in developing projects that share daily business decisions. The upshot is that business and IT collaboration is an excellent model for your organization to grow its capability and accelerate corporate ROI.
In the corporate world, modern vehicles are becoming computers on four wheels, generating a stream of performance and GPS data every mile they travel. Sensors on fleet vehicles collect data on tire pressure, hydraulics, and other vehicle components, allowing fleet managers to determine that, say, an engine is about to fail, even weeks in advance.
The challenge is that managing the massive amount of information generated by connected devices – the Internet of Things – can overwhelm IT departments. Transmitting every byte of data generated by every connected device to the cloud for processing and analysis would bring enormous back-end data costs.
The solution? In this case, the IT department of a large transportation company turned to edge analytics, a process of analyzing data close to where it is collected. Telemetric sensors in the engine, camera, and other parts of the vehicle capture streaming data and dispatch them to an IoT gateway on the vehicle itself, which can analyze the data in real time. The driver can be instantly alerted to an imminent problem, or the information can be sent to the cloud for more detailed analysis by IT teams in operations or headquarters. What’s more, the cloud can aggregate all the data and allow fleet managers to make even more strategic decisions, like managing replacement parts, setting fleet policy, and deciding which vehicles to dispatch.
The Internet of Things aims to become a vital tool fueling companies’ predictive analytics so they can make instantaneous course changes on marketing programs and other initiatives. The value of IoT data degrades quickly, though, so the insights must be gleaned and acted upon promptly. Data that indicates an engine is about to fail becomes worthless once the engine actually does fail. By processing and analyzing information at the edge, companies can turn the Internet of Things into a source of insight instead of an overwhelming fire hose of data.
As companies develop more expertise in predictive analytics, their thirst increases for gathering actionable data and more informed decision-making. They need to build capability to scale their analytics solutions to grow as the business gains knowledge on how to analyze and act on acquired data.
For example, an international confectioner wanted to ensure that its inventory, sales and marketing were aligned for the critical Easter buying season, a time when a change in consumer taste can make or break the company.
To maintain agility and contain costs, the company used a cloud infrastructure to tackle the compute-intensive extract, load and transform (ETL) process for historical data, reserving their business warehouse for analytics workloads. The tightly integrated solution allowed the confectioner to combine historical information with weekly sales data from each geographic region, permitting managers to consult interactive dashboards to identify trends leading up to Easter. As a result, managers made better decisions on which inventory, packaging, and marketing campaigns would pull consumers into stores in different parts of the country.
The outcome speaks for itself. The company boosted sales over the previous Easter season, while overall inventory was reduced. In short, the stores were stocking more of the stuff consumers wanted and it was flying off the shelves faster.
While this was an impressive immediate gain, the company also kept an eye on the future. Their approach leveraged Hadoop*, an open-source enterprise platform designed to handle a large number of data sources, data flows, and volumes. Hadoop can add nodes easily and cost-effectively as workloads and customers increase, helping future-proof the initiative. Given that the confectioner had a large number of products and was constantly expanded, it wanted a predictive analytics platform that could grow quickly with it.
Just as predictive analytics leads to performance enhancements, IT departments should design their predictive analytics projects with an eye to continual improvement, while giving deep thought to how people and machines will work together to improve decision–making.
Consider TeacherMatch*, which provides analytics tools that help school principals hire the most effective teachers for their districts. In the past, hiring has been a highly subjective process, with much more art than science. TeacherMatch* collects data from multiple sources, including standard information such as state credentials and sample lessons. The system also has an Educator’s Professional Inventory (EPI) tool that asks applicants questions on four different areas -- attributes or qualifications, cognitive ability, attitudinal factors, and teaching skills – and correlates those to the likelihood that the teacher generates a particular degree of student achievement.
By pulling these different data streams together, school principals can more easily identify the best possible candidates from a large stack of resumes, balancing the quantitative and qualitative aspects of decision-making. Critically, the system leverages machine learning. The actual student achievement – like grades and surveys of how well they liked a teacher – are fed back into the system, allowing the algorithms to be continually refined.
Predictive analytics is co-dependent on human resources, including by the skills of the IT people but also how decision makers use the information. While predictive analytics guide decision-making today, in the future they will execute those decisions, putting an even higher premium on creating and selecting the best data, as well as machine learning to constantly improve the systems and algorithms.
The need for better and better decision-making never ends. A flexible, agile infrastructure prepares companies to respond to business conditions today, and also be prepared for changing demands tomorrow.