Trying to use Google Trends for a research project for the first time. Looks totally over-hyped. Here is the correlation for the exact same search term, with two months of daily data overlapping. R-squared is a mere .5 for search for same term, same days.

I don't even see a way to pay to get more than just a "partial random sample", which is where I thought they'd go with this. Google might need to shell out a couple hundred bucks and hire a reasonably competent programmer from Upwork to pick a "more random" search sample

The basic solution to the google trends issue is just to pull a lot of extractions for the same days. Several ways to do this. Take the same search terms and dates, pull the data, run another search, and then go back to the original (sometimes gets you new data).

Another way, is to add one day to the time period you need. Google trends will get you totally different data (not just different scaling). A third way is that they appear to change the seed every day. I ended up taking 12 extractions, and then simply averaging over all 12.