Predictive analytics is the application of statistical and machine-learning models to historical data to predict future behaviours or events. Unlike descriptive BI ('what happened?') and diagnostic BI ('why?'), predictive analytics answers 'what is going to happen?'. Typical cases: demand forecasting, employee flight risk, the probability of a customer defaulting, equipment failure in the next 30 days.
Successful implementations share characteristics: they connect to concrete operational decisions (a forecast without action is an academic exercise), they rest on data of sufficient quality (garbage in, garbage out), they include continuous validation (a model that has lost accuracy needs retraining), and they respect the principles of explainability and governance.
In the INFOS portfolio, predictive AI is applied at several points: the AI layer of pplPortal supports predictions of absenteeism and flight risk in HR; predictive maintenance in an industrial context (with KORA Productivity + IIoT) anticipates equipment failures; dynamic pricing in retail adjusts prices in real time based on observed elasticity. All models comply with the AI Act — explainability, human oversight, auditable logs.