Operating System Market Share Csv Data: Dataset Structure and Field Coverage

The shape of the series matters as much as the length. At monthly granularity each series carries roughly 20 OS-version columns over about 210 months, which works out to approximately 4,000 values per platform-region combination, multiplied across dozens of platform-by-region-by-country combinations. Granularity is selectable per series: daily, weekly, monthly, quarterly or yearly.

There is no API in this record; delivery is the CSV export attached to each chart page, sitting alongside PNG and HTML as one of three published formats. That constraint sets the working sequence:

1. Pick the platform series. Desktop, mobile, tablet and console are tracked separately, so a laptop procurement model needs the desktop series, while a mobile-app team needs mobile.
2. Set the geographic scope. Worldwide, one of six regions, or an individual country - the country cut is what turns a global trend deck into a launch-market argument.
3. Choose the granularity. Daily, weekly, monthly, quarterly or yearly per OS-platform-region combination; monthly is the standard choice for multi-year comparisons.
4. Export the CSV rather than transcribing the rendered chart. One export returns the full selected history - roughly 4,000 monthly values for a single platform-region series since January 2009.
5. Re-export after the revision window closes. StatCounter revises figures for 45 days after publication, so a chart locked in the week after a period ends can drift from the settled number.
6. Credit Statcounter in the deliverable. The commercial delivery terms-SA 3.0 terms require appropriate attribution, and the share-alike clause travels with any derivative table you publish.

"OS market share" measures three different things depending on who is counting, and quoting the wrong one is the most common analytical error in this space. StatCounter weighs pageviews from more than 1 million websites. W3Techs crawls websites server-side. Valve surveys opted-in Steam players. TOP500 enumerates the world's 500 fastest machines twice a year. Five named sources cover the spread:

The practical rule falls out of the table. Client-side questions - which Windows version, which mobile OS - belong to StatCounter, because it is the only one publishing per-country cuts as downloadable CSV. Hosting-stack questions belong to W3Techs, which cross-breaks its OS families against CMS, web server and hosting technologies. Compatibility and hardware-planning questions belong to the Steam survey, whose video-card listings alone run to hundreds of models per month. And claims about Linux dominance at the high end of computing belong to TOP500, where the OS-family aggregation is computed from the full 500-system list each edition.

W3Techs OS Usage for Websites is the standard citation for the claim that Unix-family systems run most of the web. The record tracks exactly two OS families - Unix-family versus Windows - across all surveyed websites, refreshed daily with monthly, quarterly and yearly trend views. Individual sites in the crawled sample are revisited roughly once per month.

The Steam Hardware & Software Survey answers a question none of the web-measurement sources can: what the installed base looks like on consumer gaming hardware. Valve's survey runs monthly, participation is voluntary and anonymous, and it reports OS version (Windows 11 and 10, macOS, Linux) alongside CPU cores and speed, GPU model, VRAM, RAM, resolution, language, disk space and VR headsets, each with month-over-month change. Views are filterable by Combined, Windows Only, Mac Only or Linux Only.

Four blind spots recur when these sources get quoted side by side.

Population. Pageview weighting, crawled-site weighting and opt-in gamer telemetry produce different denominators for the same word, "share". A 4% reading from one panel and a 15% reading from another can both be correct.

Depth. Only StatCounter (January 2009 to present) and TOP500 (June 1993 to present) hand you years-deep machine-readable history in this slice. W3Techs' free server-side series begins 1 August 2025, and Steam keeps roughly 18 months per category chart.

Geography. StatCounter splits by country; W3Techs publishes no country split of OS usage and Steam publishes none in the main survey. Any regional OS story outside StatCounter's coverage has no second source here.

Two sibling guides extend this one directly: our walkthrough of the software end of life tracker dates the migration windows that share data flags, and the hard drive failure statistics dataset guide pairs Steam-style hardware telemetry with Backblaze's reliability record for demand forecasting. For the full source index, browse the systems-software data hub or jump straight to free systems-software datasets.