Three questions stand out. All three map to the decisions this dataset serves, and the evidence also shows what you must clean before trusting each answer. ## 1. Is Indian startup funding accelerating or slowing, and is the money concentrating in fewer, bigger deals? This is the most clearly answerable question. Deal counts fall steadily, while the median disclosed ticket rises (step 4–6 results): - **Deal count:** about 929–935 in 2015, 993 in 2016, 687 in 2017, 309–310 in 2018, 111 in 2019 and 7 in 2020. The dataset ends on 2020-01-13, so 2020 is only a stub. - **Median disclosed ticket:** $1.5M in 2015, $1.0M in 2016, $2.25M in 2017, about $3.9–4.0M in 2018, $12M in 2019 and $9M in 2020. - **Total disclosed USD:** about $8.6B in 2015, $10.4B in 2017, $5.1B in 2018 and $9.7B in 2019. Fewer deals with much larger tickets suggests a shift from many small early bets to fewer, larger rounds. It could also mean reporting coverage thinned out over time. That makes it a good question to test rather than assume. **Caveats:** - Only 2,066 of 3,044 deals (about 68%) have a numeric amount. - The 2016 total came back as `nan`, so at least one amount value is a literal "nan" or corrupt entry. The 2016 total can't be trusted until that is cleaned. - A few dates are unparseable or malformed, such as years "15" and "720" and 7 unparsed dates. ## 2. Which industry verticals and cities attract the most capital, and how has that mix shifted over time? This one directly supports the decision of where to target investment or open an office. The dataset has 822 distinct verticals and 115 distinct cities across 2,459 distinct startups. The vertical field is **not consistent over time**. 2015 alone has 724 distinct verticals across 929 deals, while 2016–2019 have only about 19–53 per year. The earliest records use free-text labels and the later ones use a much smaller taxonomy. A vertical trend analysis needs the labels grouped into broad categories first, such as consumer internet, e-commerce, fintech and logistics. City counts are also fragmented (115 distinct values), so spelling variants will likely need merging. ## 3. What does a typical round look like for a given sector, and who funds it? This is the benchmarking question for founders: how much to raise, at which round type, and which investors to approach. Every one of the 3,044 deals has an investor listed, and there are 57 distinct round-type values. Combining vertical, round type and amount would give typical ticket sizes and a repeat-investor map. **Caveats:** - The 57 round-type values likely contain spelling variants that should be merged. - The `Investors Name` field holds multiple investors per row, so it needs splitting before you can count investors. - Roughly one in three deals has no usable amount, which will bias the ticket benchmarks. ## Suggested order Start with question 1 because the data supports it today, once the "nan" amounts and bad dates are cleaned. Questions 2 and 3 depend on normalizing the vertical, city and round-type labels first.