**Short answer: deal count peaked in 2015–2016 and has fallen steadily since. Deal size moved the other way, so the trend is not uniformly "shrinking". There is also no reliable seasonality.** **Deal volume: growth, then a steady decline** (yearly query, step-3) - Deals per year were 929 in 2015, 993 in 2016, 687 in 2017, 309 in 2018 and 111 in 2019. 2020 has only 7 deals, all in January. - The monthly series shows the same shape. Deals climbed from 51 in Jan 2015 to a peak of 104 in Jan 2016, held near 80–100 through mid-2016, then slid to about 40–50 a month by late 2017. By late 2018 it was in single digits or low teens (13 in Sep, 7 in Oct, 18 in Nov, 7 in Dec). - Part of the late decline may be incomplete reporting at the tail of the dataset rather than a real collapse. The data can't separate the two. **Deal size: rising** - Among deals with a reported amount, the median ticket rose from $1.5M in 2015 to $1.0M in 2016, then $2.25M in 2017, $3.9M in 2018, $12M in 2019 and $9M in 2020. - Fewer deals are getting larger cheques, which is consistent with a maturing market and fewer small early-stage deals being reported. **Total dollars: lumpy, with no clean trend** - Yearly totals were about $16.9B in 2015, $3.8B in 2016, $10.4B in 2017, $5.1B in 2018 and $9.7B in 2019. - Monthly totals are dominated by a few mega-deals. For example, Aug 2017 was about $2.75B from only 45 deals, and Mar 2017 was about $2.26B. Totals are therefore a poor signal of trend. - The 2015 total in particular should be read with care, because it likely contains a few very large deals. **Seasonality: weak or absent** - No month repeats as a high or low across years. Deal counts rose in mid-2015 (Jun–Sep at 86–98, versus 51–62 in early 2015), but 2016 does not repeat that pattern. The dominant movement is the multi-year decline. **Data-quality caveats** - One row has the year "15" (dated 0015-07-01) and 7 rows have no date, so they fall outside the yearly trend. - Many deals have no reported amount, for example 929 deals but 652 with an amount in 2015. Dollar figures cover only the disclosed deals. - An earlier amount-parsing attempt (step-1) produced different yearly totals and NaN for 2016, which is why I used the cleaner step-3 figures. - The monthly query matched 60 rows but showed only 50, so I did not read the monthly detail past Feb 2019.