Predictive analytics defines outcomes through models to answer the question “What will likely happen?” By doing so, businesses can move beyond reactive strategies toward positive outcomes, meeting business goals and ameliorating risks. Predictive analytics probability models trigger from historical data, sensor data, and data-in-event streams both specific customer behaviors and global news (such as a company merger or an unexpected weather event). As a sub-field of machine learning, this analytics model can describe what will likely happen within seconds or even milliseconds of generating a prediction.
Use Cases Include:
- Forecasting demand and price curves
- Preventing churns
- Handling maintenance requirements
- Identifying high-risk populations (such as high-risk patients or potential murder victims)
Beware of making predictive analytics models from insufficient data sets and unconsidered variables. For example, during the financial crisis in 2008-2009, predictive analytics use cases did not consider a price drop in housing, leading to inaccurate assessments of mortgage repayments.
Other Definitions Include:
- “Simple forecasting models in which past performance” has been used to determine future outcomes. (Kimberly Nevala)
- A “sub-field of data analytics and business intelligence, which deals with an in-depth analysis of past events and forecasts of future events.” (Paramita Ghosh)
- “Forecasts by integrating data mining, machine learning, statistical modeling and other data technology.” (Georgetown University)
- Applications or systems that see patterns in time and place, giving users foresight (i.e., when and where murders increase in Columbia). (The Economist)
Businesses Use Predictive Analytics to:
- Identify primary markets.
- Handle and mitigate risks.
- Highlight areas for new revenue.
- Keep valued customers.
- Offer customized products and services.
- Control quality.
- Evaluate marketing campaigns.
Image used under license from Shutterstock.com