Each month's table has a different number of distinct values per categorical column, shown in the rendered results (step-2 and step-3 queries). The counts below are distinct values within each month's table, not across all five months combined. | Month | Rows | Company Name | Location | Sector | Stage | Investor | |---|---|---|---|---|---|---| | January | 115 | 114 | 11 | 42 | 12 | 107 | | February | 96 | 96 | 13 | 39 | 11 | 86 | | March | 98 | 98 | 17 | 54 | 11 | 90 | | April | 95 | 94 | 39 | 58 | 18 | 54 | | May | 42 | 24 | 16 | 19 | 6 | 22 | - **Company Name and Investor** are close to unique per row. Company Name is nearly one-per-row (114 of 115 in January, 96 of 96 in February, 98 of 98 in March, 94 of 95 in April). Investor is high-cardinality, with 86 to 107 distinct values in January to March. - **Sector** is the richest low-to-mid cardinality classifier, with 39 to 58 distinct values in January to April. - **Location** has few values in January to March (11, 13 and 17 distinct). April jumps to 39, which suggests less standardised location strings that month. I did not check this. - **Stage** is a compact category with 11 to 12 distinct values in January to March, 18 in April and 6 in May. - **May is an outlier.** It has 42 rows but only 24 distinct company names and fewer distinct values everywhere. The dataset notes say some May rows are misaligned or blank. The lower counts probably reflect blank or shifted rows rather than a real drop in variety, so treat May's numbers with caution. I did not compute distinct counts across all five months together. Because the tables share no join key, each month is reported separately. A combined count would need a UNION of the tables.