A 0.05% drop is well within the expected margin of error with a perfect statistical methodology and large sample size. The steam survey is opt in, it's a good rough guide but you can't read anything into month-month changes.

Even if we generously agree that it's correct, how much has steam grown? A 0.05% drop could still be more total users than the previous month.

The usage share was 0.97% at the end of 2015, 0.87% at the end of 2016 and 0.75% in the latest survey. This might still translate to a growing absolute number of Linux users, but it would be nice to see the momentum start going in the opposite direction when it comes to those percentages.