## Column classification The dataset has 23 columns. The metadata profile covers the 19 business columns, and the dataset card lists the other 4. I did not run a separate SQL check on the VARCHAR columns. **Identifier (1)** - `User_ID` is tagged `identifier` / `near_key`. It is the row key, one row per surveyed user. **Categorical dimensions (8)** - `Gender`, `User_Type`, `Education_Level`, `AI_Tool`, `AI_Purpose` and `Would_Recommend` are tagged `classifier`. The card gives approximate distinct counts: `User_Type` about 2, `Would_Recommend` about 4, `AI_Purpose` about 7, `AI_Tool` about 16, `Education_Level` about 17 and `Gender` about 12. - `Country` and `Profession` are tagged `spatial`, but they are really geographic and occupational grouping dimensions. - The distinct counts for `Gender` and `Education_Level` look high for those fields. This suggests inconsistent spellings or labels that need cleaning before grouping. **Numeric metrics** - Typed as numbers: - `Productivity_Score` (DOUBLE, 1.6–10.0, 16 nulls) - `Satisfaction_Score` (BIGINT, 3–10, 7 nulls) - `Accuracy_Rating` (BIGINT, 2–5, 15 nulls). It is flagged `categorical`, since it is a small ordinal 2–5 scale, but it is stored as a number. - Numeric in meaning but stored as text (VARCHAR): `Age`, `Monthly_Income`, `Monthly_AI_Cost`, `AI_Usage_Hours_Per_Day`, `Time_Saved_Hours_Per_Week` and `Work_or_Study_Hours_Per_Day`. These need a cast before averaging or correlating, and they may contain currency symbols or stray text. `Monthly_Income` is also flagged as personally identifying, so only aggregate it. **Dates / times (1)** - `_ingestion_timestamp` (TIMESTAMP) is the only temporal column. It records when the data was loaded, not when the survey was answered. The dataset has no event date, so time-trend analysis is not possible. **Other columns** - `Tasks_Performed` is free-text descriptive. - `_batch_id`, `_source_file` and `_source_system` are ingestion lineage metadata (card only; not in the profile result).