Lakehouse

A lakehouse stores data as open files in cheap object storage, then adds a table layer on top so it can be queried, updated and governed much like a warehouse. The aim is one copy of the data serving both reporting and machine learning, instead of shuttling extracts between a lake and a warehouse. For lifecycle teams this matters mainly when propensity or churn models need the same customer tables the campaign segments read. Query latency is often higher than a tuned warehouse.