**The strongest relationship in this dataset is between a startup's funding stage and its founding year. Stage also moves with the log of the funding amount, and founding year relates to amount only weakly.** I ran the correlations on the five monthly tables combined, 426 rows in total. | Pair | Correlation | Rows used | |---|---|---| | Founded vs. Stage (stage encoded as an ordered scale) | **-0.54** | 256 rows with a stage | | Stage vs. ln(Amount) | **+0.40** | 256 rows with a stage and 385 with an amount | | Founded vs. ln(Amount) | **-0.25** | 385 rows with an amount | - **Founded and Stage (-0.54):** Later stages go with older companies. Average founding year drops from 2021 for Pre-seed and 2019 for Seed to 2012 for Series D and 2013–2014 for Series E and F. Younger companies are raising early rounds and older ones are raising later rounds, as you would expect. - **Stage and Amount (+0.40):** Median round size climbs steadily with stage. It is about $0.5M for Pre-seed, $1.6M for Seed, $12M for Series A, $60M for Series C, and $316–364M for Series E and F. Amounts are heavily skewed, so I used the log. The correlation is moderate because round sizes overlap a lot across stages, and the "undisclosed" stage has a $23.5M median. - **Founded and Amount (-0.25):** Older companies tend to raise more, but most of this is probably because they are at later stages. It is the weakest of the three. The other columns (Location, Sector, Investor, Founders, "What it does") are text and I did not test them against these measures. Company Name is the row key, and Founded is the only year column. **Data quality limits these numbers** - **Amount** is stored as text. The most common values are "Undisclosed" (35 rows) and blank (25 rows), and some are malformed, such as "2,000,000" without a "$", "$728.100,000" or "6400000". Amount-based figures therefore cover only the 385 rows with a usable amount. - **Stage** is blank in 127 rows. It also has inconsistent spellings: "undisclosed" vs. "Undiclosed", "Preseed" vs. "Pre-seed", "Series-D" and "$Series A". A few rows have misaligned data, such as an investor name sitting in the Stage column, mostly in May. - Cleaning these would change the exact coefficients, though probably not the ranking or direction of the relationships. The charts below show median amount by stage. {"action":"final","content":"**The strongest relationship in this dataset is between a startup's funding stage and its founding year. Stage also moves with the log of the funding amount, and founding year relates to amount only weakly.** I ran the correlations on the five monthly tables combined, 426 rows in total.\n\n- **Founded vs. Stage: -0.54** (256 rows with a stage; stage encoded as an ordered scale)\n- **Stage vs. ln(Amount): +0.40** (385 rows with an amount)\n- **Founded vs. ln(Amount): -0.25** (385 rows with an amount)\n\n**What each one means**\n- **Founded and Stage (-0.54):** Later stages go with older companies. Average founding year drops from 2021 for Pre-seed and 2019 for Seed to 2012 for Series D and 2013-2014 for Series E and F. Younger companies are raising early rounds and older ones are raising later rounds, as you would expect.\n- **Stage and Amount (+0.40):** Median round size climbs steadily with stage. It is about $0.5M for Pre-seed, $1.6M for Seed, $12M for Series A, $60M for Series C, and $316-364M for Series E and F. Amounts are heavily skewed, so I used the log. The correlation is moderate because round sizes overlap a lot across stages, and the \"undisclosed\" stage has a $23.5M median.\n- **Founded and Amount (-0.25):** Older companies tend to raise more, but most of this is probably because they are at later stages. It is the weakest of the three.\n\nThe other columns (Location, Sector, Investor, Founders, \"What it does\") are text and I did not test them against these measures. Company Name is the row key, and Founded is the only year column.\n\n**Data quality limits these numbers**\n- **Amount** is stored as text. The most common values are \"Undisclosed\" (35 rows) and blank (25 rows), and some are malformed, such as \"2,000,000\" without a \"$\", \"$728.100,000\" or \"6400000\". Amount-based figures therefore cover only the 385 rows with a usable amount.\n- **Stage** is blank in 127 rows. It also has inconsistent spellings: \"undisclosed\" vs. \"Undiclosed\", \"Preseed\" vs. \"Pre-seed\", \"Series-D\" and \"$Series A\". A few rows have misaligned data, such as an investor name sitting in the Stage column, mostly in May.\n- Cleaning these would change the exact coefficients, though probably not the ranking or direction of the relationships.\n\nThe chart below shows median amount by stage.","charts":[{"kind":"bar","x":"stage","series":["median_amount"],"title":"Median funding amount by stage"}]} ```