The table `raw.startup_funding` has 14 columns, one row per reported funding deal (3,044 rows). The meanings below are inferred from the column names and the metadata profile, not from a data dictionary. **Source columns (10)** - **Sr No** (BIGINT, identifier): the row key, running from 1 to 3,044. - **Date dd/mm/yyyy** (VARCHAR, temporal): the date the deal was reported, stored as text in day/month/year form. It must be parsed before any time analysis. The card's range shows irregular values (for example a non-breaking-space character), so some dates are malformed. - **Startup Name** (VARCHAR, descriptive): the company that received funding. - **Industry Vertical** (VARCHAR, descriptive): the startup's broad sector. It has about 903 distinct values, so labels are messy and inconsistent. - **SubVertical** (VARCHAR, descriptive): a more specific niche within the sector. - **City Location** (VARCHAR, spatial): the city where the startup is based. The column name has a double space. - **Investors Name** (VARCHAR, descriptive): the investor(s) in the deal, likely several names packed into one text field. - **InvestmentnType** (VARCHAR, classifier, categorical): the funding round type, such as seed, angel, Series A/B/C or private equity. The name is misspelled in the source, and it has about 66 distinct values, which suggests inconsistent labelling. - **Amount in USD** (VARCHAR, measure): the deal size in US dollars. It is stored as text, so it needs stripping of separators and casting to a number before it can be summed or averaged. - **Remarks** (VARCHAR, classifier, text): free-text notes on the deal. **Pipeline metadata columns (4)** - `_ingestion_timestamp` (TIMESTAMP), `_batch_id`, `_source_file` and `_source_system` (VARCHAR) record when and from where each row was loaded. They carry no business meaning. **Caveats** - The profile reports zero nulls in every column. Because most columns are text, missing values may be stored as empty strings or placeholder text instead. - Only `Sr No` is a true numeric type. The date and the amount both need casting before analysis. The column list comes from the profiling query (`inspect_columns` on `raw.startup_funding`) and the dataset card.