A partition key is an access and distribution decision. Alder Supplies frequently lists the records for one customer, but a few customers produce much more data than the rest. A customer key may suit the common query while concentrating a large customer's demand.
Write down the queries, data growth, value distribution and skew before selecting a path. A unique identifier is not automatically the best key if every useful query has to search many partitions. Conversely, a convenient low-cardinality status can concentrate traffic. These tasks create local container definitions and evaluate original access patterns. The browser does not distribute documents, consume request units, enforce partition limits, or measure query fan-out. The selected key is a training design requirement, with its tradeoffs made explicit rather than presented as a universal best practice.
Balance useful queries against key skew and growth.