# Three questions this dataset can answer The dataset is 3,044 reported Indian startup funding deals with date, city, vertical, investors, round type and amount. Three questions stand out. Each comes with a caveat from what I checked. ## 1. Is funding activity accelerating or slowing, and is it concentrating into fewer, larger deals? This is the most telling question, and the yearly profile (steps 4–6) already points to an answer. - **Deal count falls steadily.** About 935 deals in 2015 (the count differs slightly between queries because of date-parsing anomalies), 993 in 2016, 687 in 2017, about 310 in 2018, and 111 in 2019. 2020 has only 7 deals, so it is a stub. - **Median disclosed ticket rises.** It goes from about $1.5M in 2015 to $1.0M in 2016, $2.25M in 2017, about $4M in 2018, and $12M in 2019. - **Total disclosed dollars are roughly flat.** They run from about $8.6B in 2015 to $10.4B in 2017, $5.1B in 2018 and $9.7B in 2019. The 2016 total came back as `nan`, so it can't be stated until the unparseable amounts are cleaned. - **What it suggests.** Fewer, bigger deals point to a market maturing and concentrating, or to a collection that thins out over time. The dataset can't tell these apart. - **Caveats.** Only 2,066 of 3,044 amounts are numeric. 2019 has 104 amounts across 111 deals, so its median is well covered. Dates also need cleaning: a few parse as year "0015" or "720", and 7 don't parse at all. ## 2. Which cities and verticals should be prioritised, and how has that mix shifted? - The data has 115 distinct city values and 822 distinct vertical values, so it can rank where deals and dollars cluster. - Distinct cities per year fall from 54 in 2015 to 29 in 2017, which hints at geographic concentration. That is worth confirming with a per-city share over time. - Distinct verticals per year are 724 in 2015 but only 19–53 in later years. The 2015 vertical labels look like free text, while later years look more standardised. - **Caveat.** Any vertical trend needs a mapping of the 2015 labels into broader categories first, or the comparison is mostly noise. ## 3. What does a "typical" raise look like by round type and sector, and who funds it? - There are 57 distinct round types and every deal has an investor field, so you can benchmark ticket size by round (seed, angel, Series A/B/C, private equity) and by vertical. - You can also build an investor-to-sector and investor-to-city map for deciding whom to approach. - **Caveats.** Investor names are likely multi-valued strings that need splitting. Round-type labels need normalising, because 57 values is more than the real number of stages. The median ticket moves a lot by year, so benchmarks should be set per period rather than pooled. ## Data-quality points to fix before any of this - About a third of deals (978 of 3,044) have no numeric amount. - Dates are stored as text, with a handful that are malformed. - One or more amount strings break the 2016 total. - Category fields are inconsistent in spelling and granularity.