The funding data has quality problems in five areas. The biggest is that `Amount` can't be summed or compared as it stands, and the May table is partly broken. **1. `Amount` is free text in many formats, with mixed units** `Amount` is a VARCHAR column. The format breakdown (step 6) is: | Format | Rows | |---|---| | `$` with comma separators, e.g. `$1,000,000` | 354 | | "Undisclosed" | 35 | | Blank | 25 | | No `$` prefix, e.g. `1,000,000` | 11 | | Mixed `.` and `,`, e.g. `$11.300,000` or `$728.100,000` (April only) | 11 | | `$` decimal with no stated unit, e.g. `$165.5` or `$84.2` | 5 | | `$` with no separators, e.g. `$60000000` | 2 | | Other/text, including `$1,40,000` (Indian-style grouping) and the non-amount value `Vaibhav Anant` | 2 | | Bare number, `6400000` | 1 | - **Unit ambiguity:** `$165.5` or `$84.2` are probably millions (MobiKwik, OkCredit, Livspace, WayCool), but the data doesn't say so. April's `$728.100,000` for Fivetran cannot be parsed reliably, because the decimal point and comma conflict. - **Possible billion-dollar outliers:** 10 rows in a single month have amounts from `$1,100,000,000` to `$5,000,000,000` (step 1). The data can't tell me whether these are real mega-rounds or unit errors. Verify them before using them in totals or averages. - **Implausible amount for a small round:** YouKraft (founded 2020, Seed stage) shows `76,000,000`, which looks suspect for a seed round. - **Blanks are not NULLs:** the null counts for `Amount`, `Stage` and `Location` are all 0 (step 2), yet 25 `Amount` values are empty. These are empty strings, so `IS NULL` checks will miss them. **2. `Stage` labels are inconsistent** - "undisclosed" and "Unknown" are used alongside blank values, and all three mean "no stage". - "Series-D" appears next to "Series D". - Several April rows have a stage of "undisclosed" even though an amount is given (for example Doceree and MoEngage). **3. `Founded` has impossible or suspicious values and mixed types** - April includes Rigi (1871), MTR Foods (1924), Philips Electronics (1929) and Hitachi (1959). These are established corporations rather than startups, and 1871 is probably a data error. - `Founded` is BIGINT in four tables but VARCHAR in May. In May, 19 of 42 rows hold non-numeric values (step 2). **4. The May table is structurally corrupted** - FirstCure Health has its columns shifted: "Noida" sits in `Founded`, and the sector and description are in the wrong columns (step 3). - There are 11 rows that are entirely blank. - May shows 18 duplicate company names, mostly the blank names. April and January have 1 duplicate each. **5. Quality varies by month** - February and March have no `Founded` problems and no duplicate names (step 2). - The mixed-separator amounts occur only in April. - Each month is a separate table, so any cross-month analysis has to clean and union them first. **Before analysing** - Normalize `Amount` to a single numeric unit. - Treat blank, "Undisclosed", "Unknown" and "undisclosed" as missing. - Drop the blank and misaligned May rows. - Cap or check `Founded` values below roughly 1990.