Understanding Windows 10 Usage Share (Premium)

Windows 10 had “no growth at all” in the last month, some tech pundits claim. But that’s just a click-bait headline, and one that confuses marketshare with usage share.

Which, of course, most people still don’t understand.

A quick recap.

Marketshare measures unit sales as a percentage compared to the broader market. It is compared each month year-over-year. So February’s sales from this year would be compared to that from February 2017.

Usage share measures usage as a percentage compared to the broader market. It is compared each month year-over-year as well as month-over-month. So February’s usage from this year would be compared to that of February 2017, and it would be compared to January 2018 and previous months. In doing so, we can see trends emerge.

Both marketshare and usage share are estimated, mostly because there is so little hard data. But that is especially true for usage. And finding reliable sources of data in either category is problematic.

For marketshare, I rely on Gartner and IDC, but I average their results because each uses slightly different methodology (especially around PC/tablet sales).

For usage, I trust Netmarketshare because I know that Microsoft does as well, and I use them primarily for web browser and OS usage estimates. But others point to StatCounter as an alternative.

Regardless of which you prefer, usage share is what they measure, not marketshare. And it’s not just because people don’t “buy” a web browser; you could argue that people do “acquire” a new browser from time to time. But that’s not what Netmarketshare and StatCounter measure. They measure usage. Period.

Anyway. This past week, Netmarketshare released its usage share figures for February 2018. (Even Netmarketshare calls this “marketshare” by the way. Don’t be fooled.) And as a few bloggers have discovered, the February numbers show that Windows 10 usage has fallen, month-over-month.

This is a meaningless data point.

Windows 10 usage share is not going to go down over time, it’s going to go up. The speed at which it goes up may vary—it was growing fast during the free upgrade period two years ago and more slowly since—but that’s about it. There almost certainly will not be dips in real-world usage, not over time, and not even month-over-month. If we do see such a month-to-month dip, it has to be a statistical anomaly. It’s not like a bunch of new Windows 7 and 8.1 PCs (or Linux PCs or Macs) just came online suddenly, skewing our understanding of where the growth is.

Here’s what happened, according to Netmarketshare.

Windows 10’s usage share in February was 34.06 percent. So a bit over one-third of all PCs used worldwide are running Windows 10.

That figure is a significant jump over the 24.48 percent usage share that Windows 10 had in February 2017, one year earlier. And you can see that the reason for this change is a drop-off in older Windows versions. It all matches up nicely.

But that 34.06 percent usage is “smaller,” slightly, than 34.29 percent usage that Windows 10 had a month earlier. It’s .23 percent smaller. Point twenty-three. So it’s more accurate to say that Windows 10 usage was flat, month over month. And it’s equally fair to point out that February isn’t exactly a big month for new PC sales. (I’ll also add that Windows 7 recorded a bigger month-over-month fall from January to February.)

Look at a year ago. Yes, Windows 10 usage did “grow” month-over-month, but only by 1 percentage point. Like .23 percent, that number represents flat growth. From February to March 2017, the growth was just .60 percent. The needle just doesn’t move much at this time of year.

So again, this is a meaningless data point. And a clickbait headline can’t hide the fact that this isn’t even worth reporting on. I would have ignored it myself had it not been for those clickbait headlines. Which I will not link to.

Windows 10 marketshare, by the way, is a soft number too. We get PC sales data each quarter, not each month, for starters. That helps smooth out these weird month-over-month anomalies. And, of course, not every PC sold ships with Windows 10, though most do. More important is the notion of Windows 10 licenses sold, a figure Microsoft will never provide. Round and round we go.

But that’s OK: From time-to-time, Microsoft provides a very useful and accurate number for Windows 10 usage (not marketshare): It reports on the number of active devices (the vast majority of which are PCs) that are now running Windows 10. At last count, this was 600 million. And I think it’s fair to say that, as of today, there are roughly 600 million Windows PCs—and users—out in the world. That’s about as “hard” a number as we’re ever going to see here.