Here is a 15-row sample spanning April 2015 to April 2019, shown in the rendered block above. It was chosen to cover different years, cities, sectors and round types. **What the sample shows** - **Grain:** each row is one reported funding deal, with a date, startup, industry, city, investors, round type, amount and remarks. Examples are Medlife (Private Equity, Bengaluru, 2019), Ninjacart (2016) and Sulekha.com (2015). - **Cities:** Bangalore, Mumbai, Gurgaon, Chennai and Jaipur all appear. - **Round types:** most rows are "Private Equity", with two "Seed Funding" rows (Dropkaffe and Kleeto). Because the sample is small, it cannot show the true mix. - **Investors:** the field is free text with several comma-separated names. Some entries, like the Accel Partners / Nandan Nilekani row for 4tigo, end in a trailing comma. **Data quality issues visible in the sample** - **Amounts are text in Indian digit grouping.** A value like `1,70,00,000` means 17,000,000. Some are `N/A`, as for Digit. The column must be cleaned before any numeric analysis. - **The columns shift for older rows.** In the 2015 rows (Razorpay, Kleeto, TimeSaverz, CreditMantri, Sulekha.com), `Industry Vertical` holds what looks like a sub-category such as "Online Payment Gateway". `SubVertical` is `nan`, and `Remarks` holds the round label, such as "Series A" or "Series C". - **Labels are inconsistent.** "E-commerce" and "ECommerce" both appear, as do "Bengaluru" and "Bangalore". Grouping by these fields without normalising would split the same category. - **Missing values are stored as the string "nan"** rather than as nulls, and `Remarks` is sometimes empty. I can run a cleaned-up view next, such as parsed amounts by year or a normalised city or industry breakdown.