All five monthly tables (January to May 2022) share the same 13 columns. Each row is one startup funding round announced in that month. The types below come from the dataset card and the column profile I pulled for the January table. **Business columns** | Column | Type | Meaning | |---|---|---| | Company Name | VARCHAR | The startup that raised funding. It is the row key, with about 107 distinct values in January. | | Founded | BIGINT (VARCHAR in May) | The year the company was founded. January ranges 1998 to 2021. | | Location | VARCHAR | The city where the company is based, with about 10 distinct cities in January. | | Sector | VARCHAR | The industry category, with about 40 distinct values in January and about 55 in March. | | What it does | VARCHAR | A free-text description of the company's product or business. | | Founders | VARCHAR | The founder names, as comma-separated text. | | Investor | VARCHAR | The investor or investors in the round, as free text. Some are blank. | | Amount | VARCHAR | The funding amount, stored as text such as `$1,000,000`. | | Stage | VARCHAR | The funding round stage (Seed, Series A to F, and so on). | **Ingestion metadata columns** These were added when the data was loaded and say nothing about the startups. - `_ingestion_timestamp` (TIMESTAMP): when the row was loaded. - `_batch_id` (VARCHAR): the load batch identifier. - `_source_file` (VARCHAR): the path of the source CSV. - `_source_system` (VARCHAR): the origin of the data. In January it is always `api`. **Data quality caveats** - **Amount is text, not a number.** Before any summing or averaging you need to strip `$` and commas and cast it. Some values are not amounts at all. In January the maximum value of the column is "Vaibhav Anant", which looks like a founder name that landed in the wrong column. - **May is misaligned.** The card notes that some May rows are misaligned or blank. That is also why May's `Founded` is VARCHAR, since a non-year value such as "Noida" appears in it. - **Investor and Stage have blanks.** The profile shows empty-string values rather than true nulls, so null counts understate the missing data. - **No joins.** The tables are independent monthly files with no detected shared keys. To analyse across months you would have to `UNION` them, and the column layout makes that straightforward.