This page  says "Total: 46.37B -41.80% over this time range."
This is a meaningless metric, and very obviously also false.
Problem is: what the script does is take the last and first count and calculate the ratio.
1) Obviously the last count is for an incomplete month, it's 80-90 % less than previous months.
2) Any trend analysis which ignores all measurements except two is painting a meaningless picture.
Lots of sophistication could be applied. A rather simple one (KISS) would be calculation of the ratio of 12-month averages for first 12 months and last 12 months (excluding instead of focusing on first and last month which could be incomplete months).
Better not compare first and last calendar year, as that could be based on 12 months, but also on 1 and anything in between. So in this case it would average for all months in 2008 vs average of Aug 2015-July 2016, ignoring Dec 2017 and Aug 2016.
Also presenting the total for all months (46 B) is quite uncommon for bar charts and likely to be misinterpreted.