December 01, 2017

NMS make major changes, and finally explain themselves.

I'm quite late posting today's OS market share numbers, in part because I honestly wasn't expecting much to change from last month. Well, was I ever wrong on that, because NetMarketShare have changed just about everything... and, more importantly, they've finally explained why:

Welcome to the new NetMarketShare.com


We completely rearchitected the product to provide more accurate and accessible data. Here's what's new:

Elimination of botnet data


As bot traffic across the web has risen dramatically, it has been a challenge to detect and remove it from our dataset. This is a critical issue since bots can cause significant skewing of data. In particular, we have seen situations where traffic from certain large countries is almost completely bot traffic. In other countries, ad fraudsters generate traffic that spoofs certain technologies in order to generate high-value clicks. Or, they heavily favor a particular browser or platform.

The primary focus of this release was to build detection methods to eliminate this traffic. We rewrote the entire collection and aggregation infrastucture to address this issue.

Please note: This dataset is separate from and replaces the legacy data. Subscribers that need legacy data please contact us for access.
The "bot traffic" explanation is... plausible, actually. Bots are a huge online problem, and for a business whose business is measuring online traffic, it makes sense that large-volume bot activity would seriously mess with their data, and their analyses.

For the tracking of OS usage numbers, of course, this basically renders every previously posted result (i.e. the "legacy data") completely irrelevant. Therefore, I will let it go, and start over fresh, much as NMS have themselves. It's important to keep some sense of the history here, through, so... let's see what NMS have on tap for history.




Annoyingly, the new table which accompanies this graph doesn't break out each month's numbers in a nice grid, but rather shows only the average for the period. Each month's numbers are only available as mouse-over tooltips, which is worse for at-a-glance comparisons but probably better for NMS's business model (which is, after all, to sell the numbers to subscribers). Also, the new NMS data doesn't go all the way back to Windows 10's launch, so we can only look at the range from May of 2016 through the end of November, 2017.

Even so, the botnet-less trend lines look rather different than the botnet-filled versions. Mousing over the data points on that graph, we can learn that Windows 7 currently holds 43.12% of the market, basically unchanged from last month's 43.05%, while Windows 10 holds 31.95% of the market, an uptick from last month's 28.86%. Interestingly, Windows 10's growth has come mostly at the expense of Windows XP, which declined from 7.64% to 5.73%.

Generic "Linux" only shows up as 1.65% of the market, down slightly from last month's 1.85%, although that doesn't include Ubuntu (which is also Linux) or Chrome OS (which is also Linux), etc. That kind of fragmentation is part of the problem that Linux has in attracting new users, though, so that seems fair enough, and at least NMS is now listing Chrome OS separately as a desktop OS, which will allow us to track its progress over time - potentially very interesting. Combining all the various versions of each OS together gives a better overall picture of how Linux-based OSes compare to Windows:


Exactly why botnets would be inflating Linux' OS market share numbers, rather than Windows', is something that I'd like to see explained in more detail at some point. Linux is famously more secure than Windows, generally, so it would make far more sense for botnet traffic to be inflating Windows' market share results, instead. I'll be keeping an eye out for a better explanation there.

In the meantime, though, we have a new, and hopefully more reliable, baseline for watching the progress of the OS market from this point on. It'll be interesting to see what happens. I'm still not expecting to see much movement for the next few months, for the reasons previously stated, but I wasn't expecting NMS to revamp their entire business in a month, either, so who the fuck knows? If nothing else, it should be interesting to see what happens.