Data mining is the practice of digging through large datasets to find patterns and relationships nobody was looking for in advance. It differs from ordinary reporting in that respect. A report answers a question you already had, while data mining surfaces a structure you did not know to ask about.
The common techniques fall into a few groups. Association rule mining finds items that occur together, clustering groups similar records, classification predicts a known label, and anomaly detection flags records that break the pattern. Most of the work happens before any of that, in cleaning, joining and shaping the data.
The boundary with machine learning is fuzzy and mostly a matter of intent. Machine learning is generally aimed at prediction, data mining at discovery. In practice the same algorithms serve both.
A concrete case: a supermarket chain runs association analysis over basket data and finds that a particular baby product and a particular snack appear together far more often than chance would explain. Nobody planned that question. The shelf layout and promotion calendar change as a result.
A found pattern is not automatically a cause. Correlations turn up in large datasets by coincidence too, so anything acted on gets tested against a fresh period before it becomes a decision.