**The strongest relationships in this dataset are among the descriptive and categorical columns. Most of the top scores are inflated by high cardinality, so treat them with caution.** The ranking comes from the normalised mutual information (NMI) results of the pairwise analysis (step-1). The dataset has no gold tables, and these columns are mostly categorical, so ordinary Pearson correlation doesn't apply. **Strongest pairs (NMI, 0–1)** - **SubVertical and year: 0.974.** This is the highest score, but it is mostly an artefact. SubVertical has 1,943 distinct values across 3,044 rows (step-3), so almost every value is unique to a year. - **Industry Vertical and Remarks: 0.888.** Remarks has about 75 distinct values, so this may reflect real structure, with certain remark values tied to particular verticals. I did not check which ones. - **InvestmentnType and SubVertical: 0.769.** - **City and SubVertical: 0.694.** - **Amount (log10 bucket) and SubVertical: 0.655.** - **Industry Vertical and year: 0.582.** - **Industry Vertical and SubVertical: 0.525.** This is expected, since sub-verticals nest within industries. - **City and Industry Vertical: 0.425.** Some verticals cluster in particular cities. - **Industry Vertical and InvestmentnType: 0.419.** **Weak or negligible relationships** - **Amount and City: 0.07.** Deal size is nearly independent of location. - **Amount and year: 0.036.** Deal size is nearly independent of year. - **Amount and Remarks: 0.112.** - **InvestmentnType and Remarks: 0.176.** - **City and Remarks: 0.173.** - **Amount and InvestmentnType: 0.267**, a modest link between round type and ticket size. - **Amount and Industry Vertical: 0.299**, also modest. **Caveats** - **Cardinality inflation:** NMI rises mechanically for columns with many distinct values. Industry Vertical has 822 distinct values and SubVertical 1,943. Pairs involving them should be read as "these columns are very granular", not as strong dependence. - **Amount data quality:** Amount in USD is stored as text, and only 2,066 of 3,044 values parse as numbers (step-3). The median parsable amount is about $1.73M. The amount-based relationships use only the parsable rows. - **Identifier columns:** Startup Name and Investors Name are 100% unique, so they carry no usable association with other columns. Sr No is just a row identifier. **Most dependable findings:** the relationships among the lower-cardinality columns are the most reliable. These are Industry Vertical, City, InvestmentnType and Remarks. Funding amount is only weakly tied to everything else.