Windows installer types and their market prevalence

Measured on 15,880 WinGet packages and 50 popular products: 56.9% install through a setup-engine EXE, 16.6% through an MSI and 1.6% are MSIX or APPX. What that means for teams planning a move to Windows 365, Azure Virtual Desktop or Intune.

Installer types in the WinGet repository
Vendor-published MSIX in popular products
MSIX share of WinGet over time
Add Remove Programs, Scoop and Chocolatey compared
MSI and MSIX in enterprise packaging
What the figures mean for packaging teams

Research paper. 14 pages. Reviewed 3 November 2026.

Research by Ryan Mangan from EfficientEther.

What this paper helps you decide

Decide where your packaging work sits by seeing how Windows software is distributed and where MSIX stands.

Windows software reaches devices through several overlapping installation technologies, and public evidence on how common each one is has mostly been indirect. That matters to a team planning a move to Windows 365, Azure Virtual Desktop or Microsoft Intune: most software in public catalogues arrives as an EXE setup, a portable app or an MSI, and few popular products publish an MSIX, so a migration plan needs a route from those formats to the packages the target platform expects, and not only a way to deploy them.

The paper measures it directly. It counts every one of the 15,880 packages in the Windows Package Manager (WinGet) community repository once, by its primary installer type: 56.9% install through a setup-engine EXE (NSIS, Inno Setup, custom EXE or WiX Burn), 24.9% are portable applications, 16.6% are MSI and 1.6% are MSIX or APPX. A check of 50 popular products, fixed by rule before any was examined, finds 7 (14.0%) with a vendor-published MSIX, 6 more (12.0%) with an MSIX in the Microsoft Store only and 36 (72.0%) with no MSIX route.

A re-count of the repository at five past dates follows the MSIX share from 0.99% in September 2023 to 1.62% in November 2026, while the catalogue tripled. Since September 2025 the rise has come mostly from the catalogue growing, not from a surge of MSIX among new entries. The paper is explicit that these figures describe software availability, not installed devices or usage. This revised edition also corrects the first edition, whose hand-picked list of nine vendors showed MSIX for six (66.7%): a list fixed by rule shows 7 of 50 products (14.0%, 95% interval 7.0% to 26.2%).

Who it is for

Application packaging and endpoint teams, end-user computing leads planning a move to Windows 365, Azure Virtual Desktop or Microsoft Intune, and MSPs that package software for customers. It suits estates of any size, because the figures describe public software and the test that matters is your own applications.

Get the paper

What you leave with

How Windows software is distributed: of 15,880 WinGet packages, 56.9% install through a setup-engine EXE, 24.9% are portable, 16.6% are MSI and 1.6% are MSIX or APPX.
How often popular products publish an MSIX: 7 of 50 (14.0%) from a list fixed by rule before checking, with six more in the Microsoft Store only.
How the MSIX share of WinGet moved, from 0.99% in September 2023 to 1.62% in November 2026, and why the recent rise is mostly the catalogue growing.
Why one figure is not enough: MSIX is 1.6% of WinGet packages, 0.3% of the applications examined by Add Remove Programs, 0.27% of Scoop manifests and 14.0% of the 50 popular products, because each source measures something different.
What the figures mean for a move to Windows 365, Azure Virtual Desktop or Intune: the packaging work sits in getting EXE, portable and MSI software into the packages the target platform expects.
What the evidence still cannot show: installed-base usage, private line-of-business software, whether declared installer types match the installers, and operational outcomes.

What you get

14 pages, a revised edition dated 2 November 2026, with six figures, comparison tables and 27 numbered references.
The method in full: a census of microsoft/winget-pkgs at one commit, a vendor check whose product list was frozen at 22:44 UTC on 2 November 2026 before any product was checked, and a re-count of the repository at five past dates.
A revised edition that replaces the first edition's hand-picked list of nine vendors with a check fixed in advance, and reports the MSIX figure as 7 of 50 products (14.0%) instead of 6 of 9.
Company-authored research that has not been peer reviewed. EfficientEther makes EtherApps Forge, a commercial interest in the question the paper asks, and the paper discloses it and states that the census, the vendor check and their scripts are retained so that others can rerun them.