Tools alone do not make a team productive. On August 9, 2000, Joel Spolsky, who went on to co-found Fog Creek Software that September and Stack Overflow in 2008, published "The Joel Test: 12 Steps to Better Code" on his blog Joel on Software (https://www.joelonsoftware.com/2000/08/09/the-joel-test-12-steps-to-better-code/ 18,518 ). It rates a software team with twelve yes-or-no questions, one point per "yes". Spolsky called 12 perfect, 11 tolerable, and 10 or lower a sign of serious problems. He added that most software organizations scored only 2 or 3, while companies like Microsoft ran at 12 full-time.
A high score does not guarantee a good product; a team can pass every question while building the wrong thing. The value of the test is speed: Spolsky noted it takes about three minutes, so you can even score a team you are interviewing with.
The questions have aged well, but the answers look different on a front-end team in 2026.
| # | Question | Typical 2026 answer |
|---|---|---|
| 1 | Source control? | Git 1,932 on GitHub 29 , GitLab 377 or Bitbucket 2,046 |
| 2 | Build in one step? | npm 2,036 run build, identical locally and in CI |
| 3 | Daily builds? | CI on every push and pull request |
| 4 | Bug database? | GitHub Issues, Jira 619 or Linear |
| 5 | Fix bugs first? | Failing checks block merges |
| 6 | Up-to-date schedule? | Roadmap and milestones tied to issues |
| 7 | A spec? | Design doc or RFC reviewed in a pull request |
| 8 | Quiet conditions? | Focus time, async chat, remote work |
| 9 | Best tools? | Fast laptops, big screens, AI assistants |
| 10 | Testers? | Automated tests plus QA and exploratory testing |
| 11 | Code in interviews? | Pair-programming or take-home exercise |
| 12 | Hallway usability? | Quick sessions with five users, preview deploys |