What is data mining?

Inside the years of invoices, orders and customer history your company already stores lie patterns nobody has gone looking for: products systematically bought together, customers whose ordering rhythm announces their departure months in advance, fine-grained seasonality that planning “from experience” misses, transactions that break the pattern exactly the way errors and fraud do. Data mining is the discipline that digs for these veins with statistical and machine-learning methods: association rules (the famous “buyers of X also take Y”), automatic segmentation of customers into natural groups, prediction from history, anomaly detection. The difference from ordinary reporting is the direction of the question: a report answers what you asked (“how much did we sell in March?”), while mining surfaces what you didn’t know to ask (“March is weak only for segment-B customers, and has been for two years”). For a small or mid-sized company, the sensible entry point is not an expensive platform but a focused project on a question with money in it — which customers are we at risk of losing? which products belong in bundles? — using the data you already have and a capable analyst.