Imagine a fashion retailer already knows what colors, cuts, and sizes customers will prefer next season. Imagine that workers on the shop floor in a production line know when and why a machine will fail before it actually happens and the production flow interrupts. Imagine an insurance company which can precisely estimate the likelihood that a potential new insurance customer will cause significant car damages in the future. All this knowledge would help companies to make even today good decisions for the future. But how to predict the future?
In the past, individual managers with experience and their intuition for trends and industry developments made the difference. Today, modern data analytics bring companies closer to that goal. By using historical data and statistical models such as regression analyses, companies can better forecast future developments. These data-supported scenarios can be used to optimize decisions in a variety of fields such as product program design, pricing, production maintenance, or customer acquisition. In Moneyball, a famous Hollywood movie based on a real story, the successful baseball team Oakland Athletics is formed out of players who scored badly based on traditional baseball metrics but exceptionally good based on predictive statistical analyses.
What makes predictive analytics a very hot and exciting topic in these days is a general trend called “digitalization” accompanied by some recent developments which allow companies to enlarge data opportunities more than ever before:
- Firms collect more data than in the past from a variety of sources such as their online shopping websites, internal accounting systems, or modern production machines equipped with sensors.
- Most of these data is already machine-readable and can be exploited with modern analytics software; new techniques even allow to gather and structure unconventional data from sources such as website texts, audio podcasts, thermal sensors, or videos, and make them analyzable for statistical analyses.
- Data can be acquired, stored, and analyzed in real-time due to more powerful IT infrastructure than in the past: for instance, in online shopping, customers immediately get new product offers based on what they just bought or showed interest in; online retailers predict real-time out of shopping data customer’s creditworthiness and accordingly offer different ways of payment (e.g., pre-payment or by invoice) to different customers.
- Finally, university graduates are better educated to develop and use statistical models for prediction than in the past and new and specialized job profiles such as business analysts or business intelligence managers have been developed in companies.
However, as with any technique, there is a risk. Predictive analytics can only work if historical data is available in sufficient quality and the past has predictive power for the future. For many day-to-day business decisions, this is fortunately the case. However, disruptive changes in industries are hard to detect and historical data can even be misleading here. The success of Henri Ford’s Model T or the coffee-to-go trend supporting Starbuck’s high growth would have been hard to predict by historical customer habits. In these areas, entrepreneurial intuition and early test market research will keep its importance.