The raw table `raw.AI_Usage_and_Impact_on_Students_and_Professionals` has 23 columns, shown in the column listings I pulled. One row is one surveyed student or professional. The meanings below are inferred from the column names and the semantic roles in the catalog, since there is no data dictionary. **Identifier and descriptive** - `User_ID` (VARCHAR): unique ID for each respondent. - `Tasks_Performed` (VARCHAR): free-text description of what the user does with AI. **Demographics and location** - `Age` (VARCHAR): respondent's age. - `Gender` (VARCHAR): self-reported gender. The card says it has about 12 distinct values. - `Country` (VARCHAR): country of residence. - `User_Type` (VARCHAR): student or professional (about 2 distinct values). - `Education_Level` (VARCHAR): highest or current education level (about 17 distinct values). - `Profession` (VARCHAR): occupation or field. - `Monthly_Income` (VARCHAR): income per month. This is personal information, so it should only be used in aggregate. **AI usage** - `AI_Tool` (VARCHAR): the AI product used (about 16 distinct values). - `AI_Purpose` (VARCHAR): the main use case, such as coding or content creation (about 7 distinct values). - `AI_Usage_Hours_Per_Day` (VARCHAR): hours of AI use per day. - `Monthly_AI_Cost` (VARCHAR): what the user spends on AI each month. - `Work_or_Study_Hours_Per_Day` (VARCHAR): total daily work or study hours. **Impact and ratings** - `Productivity_Score` (DOUBLE): self-rated productivity, ranging from 1.6 to 10.0, with 16 nulls. - `Accuracy_Rating` (BIGINT): rating of how accurate the AI's output is, ranging from 2 to 5, with 15 nulls. - `Satisfaction_Score` (BIGINT): satisfaction with the tool, ranging from 3 to 10, with 7 nulls. - `Time_Saved_Hours_Per_Week` (VARCHAR): hours saved per week through AI use. - `Would_Recommend` (VARCHAR): whether the user would recommend the tool (about 4 distinct values). **Ingestion metadata** - `_ingestion_timestamp` (TIMESTAMP): when the row was loaded. - `_batch_id`, `_source_file`, `_source_system` (all VARCHAR): load lineage. **Typing caveat:** only `Productivity_Score`, `Accuracy_Rating`, `Satisfaction_Score` and `_ingestion_timestamp` are typed as numbers or timestamps. `Age`, `Monthly_Income`, `AI_Usage_Hours_Per_Day`, `Monthly_AI_Cost`, `Time_Saved_Hours_Per_Week` and `Work_or_Study_Hours_Per_Day` are stored as VARCHAR even though they are numeric measures. They need cleaning and casting before any math, such as stripping currency symbols. Apart from the three null counts above, every other column has 0 nulls.