Confirming something that I'd suspected for some time now, OS and browser tracking site NetMarketShare confirmed the obvious today:
IMPORTANT NOTICE:
After 14 years of service and being used as a primary source in tens of thousands of articles and publications, we are retiring NetMarketShare in its current form. October, 2020 is the last month of data. All billing for existing accounts has been stopped. All outstanding balances are being refunded.
Why? An upcoming change in browsers (https://github.com/WICG/ua-client-hints) will break our device detection technology and will cause inaccuracies for a long period of time.
In addition, we have focused on bot detection and removal as a key part of the quality control process. It is the most complex part of our codebase. As time has gone on, it has become increasingly difficult to manage this process. So, instead of accepting increasing levels of inaccuracy, we thought it would be a good time to call it a day.
Well... That escalated quickly.
For the record, I've been convinced for some time now that NMS's "bot detection and removal" routines were basically under-counting Linux users, and favouring Windows - not in NMS's data, but in NMS's analysis. Linux user numbers kept trending upward over a period of several months, only to suddenly correct downward; and then trend upward for several more months, only to suddenly correct downward again; etc.; etc.; which always looked more like NMS was massaging the numbers to align better with their expectations, than like any sort of genuine measurement of the OS market's actual motions.
And
no, the idea that Linux-based botnets are even a thing doesn't make any
sense at all. Linux is famously more secure than Windows, to such a
degree that anti-virus software for Linux isn't even a thing, so if there
are any "botnets" that need to be "corrected" for, shouldn't that be reducing
the number of Windows systems in the analysis, not Linux ones?
This week, though, NMS finally confirmed that they have indeed been fighting a battle against increasing inaccuracies for some time now, and losing. And also that the makers of web browsers are winning the war against data-scraping and privacy invasion, successfully preventing our metadata from being harvested without our knowledge or permission. Victory!
Flawed as they clearly were, though, NMS were finally able to admit that what they did was not working, and not only stop billing customers for increasingly bad data, but even refund customers who had already paid for data and analyses they'll now never receive. So... good job, NetMarketShare. I, for one, actually will miss you, warts and all.
It does make me wonder what StatCounter are doing differently, though, if anything. I mean, they're in exactly the same business as NMS, and producing exactly the same kind of reports; it's only their data sets and analytical methodology that differed. If NMS, who were the industry leader in this kind of thing, couldn't make it work anymore, can StatCounter realistically claim differently?
After spending way to much time tracking OS market share statistics that I always suspected were junk, I will finally be deleting the Google Sheet that I was using for the purpose. It is with mixed emotions that I give you the very last iteration of my OS market share trend tracker:
Those sure are some flat line, eh?. I spent waaaayyyy too much time on this... and yet, somehow, I regret nothing.