## Summary The dataset has a time axis, `Date dd/mm/yyyy`, which I grouped by month. It shows one long rise and decline in deal counts, a handful of very large one-month funding spikes, and several breaks in the data itself. ## 1. Deal volume: peak in 2015–16, then a long decline - **Rise:** monthly deals went from about 51–62 in Jan–May 2015 to 86 in Jun 2015 and 96–98 in Jul–Aug 2015. - **Peak:** January 2016 was the busiest month with 104 deals, followed by Feb 2016 with 100 and Jun 2016 with 98. - **Decline, 2016–17:** the count fell from about 70–100 per month in 2016 to the 40s–70s in 2017. Notable dips were Jul 2016 (64), Feb 2017 (48) and Jul 2017 (43). - **Decline, 2018:** it stayed in the 30s until mid-2018, then dropped to 23 in May 2018, 13 in Sep 2018, and 7 in both Oct and Dec 2018. Nov 2018 was a brief bounce at 18. - **Tail:** Jan and Feb 2019 each had only 3 deals. ## 2. Funding amount: a few mega-deal spikes, not a trend Monthly totals are dominated by outliers, and the deal counts in those months were not unusual. - **Jul 2015:** the plain query shows a total of about $9.8B, far above any other month. Other query variants gave about $1.5–1.6B for the same month. The spike therefore depends on a handful of very large or possibly mis-parsed amounts, so treat it with caution. - **Aug 2017:** about $2.75B from only 26 deals with an amount, with a median of just $0.82M. This is clearly driven by one or two giant deals. - **Mar 2017 and May 2017:** about $2.26B and $1.79B respectively. - **Feb 2018:** about $1.21B. - **Sep 2015 and Aug 2015:** about $1.47B and $1.06B. - **Quiet months:** Dec 2016 ($187M), Jul 2017 ($168M), May 2018 ($106M), Oct 2018 ($53M) and Dec 2018 ($58M). ## 3. Deal size is rising while volume falls - **2015–2016:** the median ticket was typically $0.4–1.6M. - **2017–2018:** it rose to about $2–4.8M, for example $3.8M in Dec 2017, $4.55M in Jan 2018 and $4.9M in Aug 2018. - **Late 2018 and early 2019:** the median was $30M in Sep 2018, $9.6M in Oct 2018, $22M in Jan 2019 and $113M in Feb 2019. These months have only 2–11 priced deals each, so the figures are noisy and likely reflect thin coverage. The pattern is fewer, larger, more selective deals over time. ## 4. Breaks and data-quality issues - **Bad dates:** one record has the year parsed as `0015` (July), which is almost certainly a typo for 2015, and 7 records have an unparseable `-00` month (total about $10.2M). - **Missing month:** there is no March 2019 in the monthly results. - **Different reporting style after Feb 2019:** monthly deal counts run 3–17 and the totals include fractional amounts, such as Apr 2019 at about $115.79M. The source and reporting style appear to change here. - **Large 2019–20 months:** Aug 2019 had about $4.5B from 17 deals, Nov 2019 about $2.06B from 15 deals, and Oct 2019 about $630M from just 3 deals. The data ends in Jan 2020 with 7 deals, which is likely a partial month. - **Missing amounts:** many deals have no amount, for example 38 of 79 in Apr 2016 and 25 of 43 in Jul 2017. Total-dollar figures therefore understate activity, especially in 2016–17. - **Query variants disagree:** some parsing variants returned `nan` for Jul 2016's total and different Jul 2015 totals. The straightforward parse (step 0 and step 2) is the one I rely on, and the mismatches suggest a few malformed amount strings. ## Takeaway Deal counts rose to a peak in late 2015 and early 2016 and then declined steadily. By late 2018 and early 2019 counts were in single digits, which looks partly like thinning coverage rather than a real collapse. The dollar spikes (Jul 2015, Mar, May and Aug 2017, Feb 2018, Aug and Nov 2019) are single-deal artefacts, so use medians and deal counts rather than monthly totals to read the underlying trend.