How many people doesyour stack rest on? — free bus factor check for a list of public GitHub projects
List the GitHub projects your business runs on. For each, we show how few people made half of last year's commits.
Start from a stack
frappe, erpnext, hrms, India Compliance, bench, MariaDB, Redis.
Projects in this check
frappe/frappe
frappe/erpnext
frappe/hrms
resilient-tech/india-compliance
frappe/bench
MariaDB/server
redis/redis
Also try
7 projects · about 14 of your free GitHub requests this hour · about 13 MB to download
Nothing you type is sent to us. Your browser asks GitHub directly. GitHub gives you 60 free requests an hour, and a project costs 2.
Sample reading · recorded 14 Oct 2026
In 1 of 7 projects, one person made more than half of last year's commits.
Thinnest link: resilient-tech/india-compliance. One person made 59% of its commits.
Projects read
7 of 7
One person
1
Concentrated (2 or 3 people)
3
Spread out (4 or more)
3
What to do
One person is a fact to plan around, not a fault. Decide who you would turn to if that person stopped: pin the version, keep a copy of the source, watch for a fork, or pay for support.
Docs, CI and chore commits count here. If a number surprises you, open that project in the health check: it counts code commits only.
Recorded from GitHub's public data, so nothing was requested from your browser. Press Check stack to read it live, or pick another stack or add a project.
Project by project
Counts every commit by a person in the last 52 weeks, docs and CI included.
CHAOSS names this metric and sets the 50% line. The three words, for 4 or more people, 2 or 3, and 1, are ours, the same as the health check's, and untested against projects that later died.
What it counts. The share of the last 52 weeks' commits made by the single most active person.
The last 52 weeks, from the week of the reading back. The week now counts.
A year of commits, largest share firstDrawn, not a project
↑ Half of the year's commits
Each block is one person's commits in the year, largest first. A drawing of the rule, not a project's numbers.
What it counts. The same share for the three most active people together.
The last 52 weeks, from the week of the reading back. The week now counts.
A year of commits, largest share firstDrawn, not a project
↑ Half of the year's commits
Each block is one person's commits in the year, largest first. A drawing of the rule, not a project's numbers.
What it counts. The fewest people whose commits add up to more than half. This is CHAOSS's Contributor Absence Factor, once called the bus factor.
The metric is CHAOSS's: the smallest number of contributors behind half of all contributions. The lines are ours.
A year of commits, largest share firstDrawn, not a project
↑ Half of the year's commits
Each block is one person's commits in the year, largest first. A drawing of the rule, not a project's numbers.
Spread out: 4 or more people. Concentrated: 2 or 3. One person: 1. Read only when at least 20 commits count.
Merge commits (GitHub leaves them out of these statistics), apps GitHub marks as bots, accounts whose names end in -bot or -robot, and three AI-agent accounts: claude, codex and cursoragent. The two name rules are ours and will go stale.
The window
The last 52 weeks, from the week of the reading back. The week now counts.
Do it by hand
Each project's contributors page on GitHub, then Insights > Contributors, set to the last year.
In a study of 133 popular projects, 65 % had a truck factor of two or less: the work rests on very few people.
What this check answers
How few people make most of the commits in each project your business runs on.
Your browser asks GitHub directly, and nothing you type is sent to us. A low number is a reason to watch a project, not to drop it: many sound projects have few maintainers.
What it rests on
GitHub's contributor statisticsLast year's commits per person, read in your browser.
The workingEach project's three numbers and the stack summary are shown with how they were counted.
What a count does not tell you
Six things a commit count cannot see. Read them before you act on a thin project.
Commits are not knowledgeA person who reviews, releases or answers questions may commit little and carry a lot.
Docs and CI countA project with one busy docs writer reads thinner than its code is. The health check counts code only.
Companies are not readFive people from one firm are five people here. CHAOSS has a separate metric for that, the Elephant Factor, and it needs each person's employer.
Accounts are not peopleOne person with two accounts counts twice. A team on one shared account counts once.
GitHub's list is cutFor busy projects it returns the top 100 to 500 people by commits ever. Anyone left out made very few.
Money and controlWho pays the maintainers, who owns the company, whether the licence can change: not read.
Who made this
We implement ERPNext and Medusa for clients and support Odoo, and earn from that, so we are not neutral about them. Each is read by the same rules as any other project, and every request is listed so you can check us. We are Frappe specialists; how we stay independent is in this essay.
The two stacks list what each vendor's own documents or manifests name, plus the HR app and the India GST app our own pages name. They were chosen before any count was read, and a project one person carries stays on the list.
When the answer is a project
When a thin project carries your business, the answer is support you can call. These are the next steps.
Talk through the thin spots
Bring the result; we say which projects we would watch and how.
Support and AMC
Someone answerable for the open-source software you run.
The one most people ask is answered first. Type to find the rest.
What is a bus factor, and what do you count?
The bus factor is how many people a project could lose before it stalls. Nobody can read that from outside. We count something close that GitHub shows: the fewest people whose commits add up to more than half of the last 52 weeks. CHAOSS calls it the Contributor Absence Factor. It counts commits, not knowledge, and not who pays whom.
The count
Not by itself. A project one person carries can be well kept. Ask what you would do if they stopped: pin the version, keep the source, watch for a fork, or pay for support. A study of 1,932 popular projects found 16 % were abandoned, and 41 % of those survived because new core developers took over (Avelino et al., 2019).
This page counts every commit by a person, exactly. The health check counts code commits only, from a sample of up to 200. Docs, chore and CI commits count here. On 14 October 2026, 75 of 92 sampled commits by Medusa's most active committer were docs or chore, so Medusa reads thinner here. Open any project in the health check to compare.
We leave out merge commits, apps GitHub marks as bots, accounts ending in -bot or -robot, and three AI-agent accounts. Those name rules are ours and will go stale. We miss an agent or bot under another name, one person with two accounts, and commits GitHub can't link to an account, which count as one author each. An agent's commit can't be tied to whoever ran it.
The names are GitHub usernames from GitHub's public data, the same you see on each project's contributors page. Naming them says whom a project leans on. We show no emails and link each name to its GitHub page. CHAOSS warns that health metrics can raise privacy questions; if you want a name off this page, tell us.
Using it
No. Your browser asks GitHub for the numbers, and our server never sees your list. We add none of it to an analytics event, and the box and the result are hidden from session recording. GitHub sees your internet address and the projects, as on any visit to github.com. A click on a link that leaves this site may be logged as an outbound click, as on every page.
GitHub lets an anonymous visitor make 60 requests an hour, counted by internet address, so colleagues on one office network share them. A project costs 2, a few more when GitHub is still counting. The page asks GitHub how many are left before it starts, and shows when they return. We could raise the limit only by asking you to sign in to GitHub, and we won't.
More tools for this job
Each answers one more question, free and with no sign-up.
We use cookies and similar technologies for analytics, advertising and session replay so we can see which pages help you and which do not. Your choice is stored on this device only and you can change it at any time from the privacy page.