**The most surprising finding is that a tiny number of mega-deals account for most of the money. One of the biggest looks like a data error.** **1. Funding is extremely concentrated (step-1 query)** - 2,065 deals have a usable amount, totalling about **$38.06B**. - The **median deal is only $1.7M**, so the average deal of about $18.4M is roughly 11 times the typical one. - Just **73 deals of $100M or more (about 3.5% of deals) make up 63.6% of all dollars**. The other roughly 2,000 deals share the remaining 36%. - Any "average round size" or "total market size" figure from this data mostly describes a handful of outliers, not a typical startup. **2. The largest single record looks implausible (step-0 query)** - **Rapido Bike Taxi, Series B, 27/08/2019, is recorded at $3.9B.** That is the biggest deal in the dataset, about **10.2% of all funding on its own**, and it is larger than Flipkart's $2.5B private-equity round. - A Series B for a bike-taxi startup at this size is almost certainly a unit or entry error. I did not verify it against an outside source. - If it is wrong, the headline totals are overstated and the concentration picture shifts. The remaining large rounds are mostly Flipkart and Paytm private-equity deals ($0.7B–$2.5B). **3. Many rows have no amount at all** - Only 2,066 of 3,044 rows (about 68%) report an amount (step-2 query), so roughly a third of deals can't be sized. - Missing amounts are probably not random, since undisclosed deals tend to be a particular kind. That makes the totals above a biased view. - Step-1 counts 2,065 amounts and step-2 counts 2,066. The one-row difference is likely a single entry that doesn't parse as a number. The step-0 output shows a "Drums Food" row with a missing amount and round type. **Caveats** - Step-2's total, maximum and share columns came back as NaN, so I rely on step-1 for the totals and shares. - The dataset card shows `Amount in USD` as text, so the cleaning rules behind step-1 affect these figures. **Takeaway:** don't read this data as "typical startup funding is $18M". The typical deal is about $1.7M, and the dollar total is driven by a few dozen very large and possibly mis-keyed records.