Every number on this site is derived from data in the electron-survey repository and regenerated daily. No figure is hand-entered.
The starting point is electron/electron-apps,
the community registry of Electron applications, pulled in as a submodule and
flattened to one YAML file per app in data/apps/. Entries that are clearly
gone — dead repositories, dead download links — are marked dead and excluded.
Detection is layered cheapest-first, and the first method that produces an answer wins. Each app records which method resolved it, shown on its page.
package-lock.json, yarn.lock,
pnpm-lock.yaml, or package.json — in that order of confidence, because a
lockfile pins an exact version while package.json only gives a range
(resolved against the stable release list).electron<N> dependency, which gives the major line for free.1815 of 2053 live apps (88.4%) currently resolve to a version. The rest are overwhelmingly proprietary apps whose installers aren’t fingerprintable.
Every app page carries a How we know block naming the exact artefact the version was read out of — the installer filename for a binary fingerprint, the lockfile for a source detection, the Arch package for an AUR inference — with a link to the artefact itself so the reading can be checked.
Not all of it is equally strong, and the pages say so:
Provenance was added after much of the data was collected. Where it could be rebuilt from what was already stored it has been, and those records are marked as reconstructed. Binary fingerprints can’t be rebuilt — which artefact answered wasn’t recorded at the time — so they fill in as apps are re-fingerprinted.
Each Electron release pins one exact Chromium build; the mapping comes from Electron’s own release index. An app shipping Electron 30 is running Chromium 124 whether or not its maintainer thinks about browsers at all.
Days behind is the number of days since stable Electron first shipped a Chromium major newer than the app’s. That date is when the app’s Chromium line stopped being where new fixes land — everything after it is a fix the app cannot get without a major upgrade. It is deliberately a floor, not a CVE count: the true exposure is at least this long.
Every CVE recorded against cpe:2.3:a:google:chrome is pulled from the
NVD API —
7718 of them — along with the Chromium version range
each one affects and its CVSS base severity.
A CVE counts as open in a given build when that build falls inside the CVE’s affected range: it was fixed in some later Chromium, and the app ships an earlier one. Both ends of the range are honoured, so a bug that only ever affected Chrome 70–80 is not charged against Chrome 120.
Two things this is not:
nodeIntegration settings are, and which subsystems it actually uses. An open
CVE is a missing fix, not a demonstrated exploit path.Counts are cached in data/cves.yml, keyed by both Chromium build and Electron
release, and refreshed with the rest of the data.
Colour comes from endoflife.date’s support status for the app’s Electron major:
Because the rule is relative to today, every app drifts toward red on its own as new Electron majors ship. Nothing here needs manual re-grading.
Ranking by “importance” needs a usage signal, and the only two open, comparable ones are:
Each app keeps the stronger of the two, not a sum or weighted blend, so a
Linux-only or macOS-only app isn’t punished for missing the other channel. Both
signals are stored on the app itself, rounded down to an order-of-magnitude
bucket (1,234 votes is recorded as 1000), so day-to-day count drift doesn’t
churn the data.
The most-used apps — those clearing roughly the top-10% bar on either channel
(≥ 25 AUR votes or ≥ 7,500 Homebrew installs in a year) — are flagged
homepage: true and lead the site’s Featured list. Everything else is still
tracked and appears under All apps; there are no fixed tiers.
The two channels overlap surprisingly little — AUR skews Linux-enthusiast, Homebrew skews macOS-mainstream — which is exactly why both are needed.