**The most surprising finding is that this isn't a clean ledger of distinct funding rounds. The same companies recur across months with wildly different amounts, and some of the "amounts" look like data-entry or conversion artefacts rather than real rounds.** **Pine Labs appears in four monthly files with four unrelated amounts (step-1):** - January: $20,000,000 - February: $150,000,000 - March: $50,000,000 - April: $1,200,000,000, with the stage entered as "undisclosed" A real company wouldn't plausibly raise $20M, $150M, $50M and then $1.2B in four consecutive months. The April figure is 60 times the January one. **Other companies show the same pattern between early-year months and April (step-1):** - **DealShare:** $45M in February and $624.8M in April, both labelled Series E. - **Chargebee:** $250M in February and $468.2M in April. - **Xpressbees:** $300M in February and $577.6M in April, both Series F. - **Livspace:** $180M in February and $431.6 in April, both Series F. - **Mamaearth:** $52M in January and $110.800,000 in April. The same stage is repeated while the amount changes, so these look like re-reports of one round rather than new rounds. The April numbers also use malformed separators such as `$110.800,000`, `$624.800,000` and `$431.6`. That suggests a different source or a unit or currency conversion applied inconsistently, but the data doesn't say which. **Other signs the data is unreliable:** - `Amount` is free text. "Undisclosed" is the single most common value (35 rows) and blank is next (25), so roughly one in seven rows (60 of 446) has no usable amount (step-0). - NowPurchase appears twice in January with an identical $2.4M Seed round, and Securden appears twice in April with the same $11.7M written two ways (`$11.700,000` and `$11,700,000`), so these are straightforward duplicates (step-1). - One group of 19 rows has a blank company name and is almost entirely empty, apart from a "May:6400000:$Series A" fragment. These are the misaligned May rows the dataset card warns about. **Implication:** summing `Amount` across the five tables would badly overstate funding, because of double-counted companies, inflated April values and unparseable text. Any total needs de-duplication by company and a cleaned numeric amount first. **Caveat:** I can't tell from this data which of the conflicting amounts is correct. I'm flagging the inconsistency, not choosing a true figure. My check of how widespread the April inflation is (beyond these examples) did not return a result, so I haven't quantified it.