Businesses have had a good, long run of succeeding based on instincts and intuition. As data-driven decisions come into vogue, advocates are making convincing arguments that the data knows more than humans do. The argument has merit, to a point.
Humans have limitations, especially when facing huge amounts of data. For instance, we’re not adept at remembering lists with more than seven things on it. Even facts we once knew, like last year’s Oscar winners or the champion of the 2016 Super Bowl doesn’t remain with us. Our brains have thresholds that prevent us from seeing patterns and nuances in large data sets like those in big data. Among analytics advocates, predictive analytics combines the best of machine learning and humans. It can do what our brains can’t, and it can find the data correlations and causations that lead to better business outcomes. But it needs the touch of someone who is business-savvy.
Consider this scenario; the board at a high-tech company is adamant that executives do a better job retaining its developers and engineers. Lots of funding has gone into hiring personnel but high turnover is preventing the company from meeting software development goals. Looking for an answer, the CEO takes this problem to the lead data scientist who creates an algorithm in the predictive analytics software and applies it to the data.
In short order the software reports back this correlation: Most employees quit on Mondays. The company cafeteria serves sushi on Mondays. The recommendation is to change the Monday menu. Clearly, even though the data shows a correlation, it’s not causation. Left alone the machine learning software could flounder forever continually deciding a new lunch plan; it will take a human mentor to re-direct the algorithm.
A human knows that employees come back after a weekend of mulling life goals and give notice. That intuition will quickly reveal that lunch-time sushi isn’t relevant to the problem and reset the algorithm to find insights and recommendations that make sense.
Reports loaded with correlations, causations, clusters, and associations
Predictive analytics with machine learning is the intersection between humans and analytics. Take the usual BI report that develops when an executive asks a data analyst to review supplier costs across three regions during the past three years. After a week or more of waiting, the analyst submits a barebones report of color-coded price points across vendors, and it’s not particularly useful to the executive. Someone on staff could have guessed which suppliers were most expensive. Besides, the week-plus time gap is completely out of step with today’s fast work pace.
That’s not the only thing out of step. Companies are analyzing historic data, and basically looking in the rearview window at what’s behind them and not anticipating what will happen tomorrow, next week, or next year.
Predictive analytics with machine learning addresses the rearview mirror problem; it analyzes both historical and real-time data to make predictions about what will happen. It also provides useful answers and real insight faster. When business analysts can interact directly with the data and look for nuances, they can create an information-packed report, and the turnaround is days or hours instead of weeks.
CTG puts the brakes on employee turnover
Covenant Transport Group unraveled why it was having high driver turnover rates using predictive analytics. With insight from experts in the HR department, CTG built multiple models based on different driver types, driving time spans, driver actions, and third-party job board data.
Predictive analytics with machine learning combined these variables and ran algorithms to identify drivers with a high potential to leave the company. At the same time, it made recommendations to CTG managers about how to respond to at-risk drivers. In less than six months the company reduced driver turnover by 15%, and it recouped its investment in the first year.
Humans working in concert with predictive analytics addressed a persistent business problem. Human instinct is to look at salary and time off; machine learning delved deeper in the data to identify the nuances, patterns, and relationships unique to CTG. As the workforce changes, machine learning will adjust to new data, updating its recommendations to CTG so the company can keep pace. Together, humans and predictive analytics closed the gap between big data and decision making, leading directly to better business outcomes.
Mike Flannagan is Senior Vice President of SAP Analytics and serves as Global Head of Market Strategy for SAP Leonardo, SAP’s Digital Innovation System. Mike is active in the startup community, as an angel investor, mentor and board member. Prior to joining SAP, Mike served as an advisor to Machine Learning Startup DataRPM (acquired by Progress Software) and was Corporate Vice President and General Manager of the Data & Analytics Software Group at Cisco Systems. Mike co-founded ChallengeUp, an IoT focused startup accelerator, with executive from Intel, Deutsche Telekom and Cisco. Mike holds multiple patents, is a published author, and a frequent speaker at industry events. He earned his MBA from Auburn University, where he now serves on Advisory Boards for the College of Business and the Graduate School.