Using the profiler's "categorical" label, 18 columns qualify. The counts below are approximate distinct counts from the profiler (`approx_distinct_count`); the exact `distinct_count` field was empty. The result is shown in the rendered block above. **Low-cardinality true categories** - `Surface` has about 4 values in each of `raw_kaggle`, `return_kaggle` and `serve_kaggle`, which matches the four surfaces (hard, clay, grass, carpet). - `Rd` (round) has about 14 values in all three fact tables. - `return_kaggle.against` has about 1 value, so it is effectively constant and carries no information. - `players_tournament_man_.year` has about 24 values, covering the years 2002–2024. **Player names** - `Name` has about 491 values in `raw_kaggle`, `return_kaggle` and `serve_kaggle`. - This figure is approximate and looks high compared with the 462 rows in `players_man_`, so treat it as an estimate rather than an exact player count. **Numeric measures that the profiler tagged as categorical** - In `raw_kaggle`, `1SP` has about 189 values, `2SP` about 134, `vA` about 62, `Aces` about 54 and `DFs` about 29. - In `serve_kaggle`, `Bpsvd` has about 236 values and `Df%` about 231. - These are numeric statistics, not categories. `Bpsvd` and `Df%` are even stored as text. Treat them as measures, not dimensions. The highest-cardinality fields the profiler did not label categorical are `Tournament` (about 3,963), `Score` (about 10,386) and `against` (about 252,129 in `raw_kaggle` and `serve_kaggle`). They behave like identifiers or free text rather than categories.