SaaSBenchmark
corpus growing — 2,020 codebases methodology v8.5.2

Everyone vibecodes. Nobody measures. We do.

?Your code is deleted the moment the scan finishes — we never store it or ask who you are. Only the anonymous numbers stay, and those grow the corpus. Code never stored

Your code, measured
against 2020 real SaaS codebases

Six independent instruments scan your repository in an isolated sandbox with no network access, then place it on the actual distribution of previously measured production projects — not a synthetic grade.

Sample #7F3A2C received — zero-network sandbox spinning up 0 / 6 instruments
clocgitleaksjscpdlizardeslintosv-scanner
cloccounting lines, detecting language
gitleaksscanning for leaked credentials
jscpdchecking for duplicated code
lizardmeasuring cyclomatic complexity
eslintrunning lint ruleset
osv-scannerchecking dependencies against OSV.dev
verifying sandbox deletion…
0repos scanned
0lines measured
0leaked secrets found
0vulnerable pkgs found
0completed measurements
Overall score
0 / 100
static score Security + Maintainability only. Performance and best-practices need a live URL — they arrive with dynamic scans.
Where you land
0th percentile
vs 567 JS samples ?Where your score ranks among every JavaScript codebase in the corpus. The dot sits at your percentile — the shaded area is the share you scored better than.

This isn’t the same number as your score. The score (on the dial) is graded against fixed thresholds; the percentile is where that score falls among real projects.
0 25 50 75 100 92 THIS SAMPLE
← worse code qualitybetter code quality →
67% scored lower this sample — score 92
What you get
Sampleno account
  • Watch a full scan, start to finish
  • See the score & where a sample lands
  • No signup just to look
Free accountthe whole thing
  • Scan your own repo — your score & percentile
  • A shareable scorecard page + a README badge
  • AI fix suggestions + the full per-file breakdown
  • Scan history & a re-scan on every push
Live-URL scancoming
  • Performance, measured by Lighthouse
  • Best-practices, accessibility, SEO
  • The two categories a static scan can’t see
Honest by construction — every number is actually measured, or shown as not-measured with a reason — never guessed, never defaulted to zero.
Tools: cloc · gitleaks · jscpd · lizard · eslint · osv-scanner — a percentile only counts against samples measured with the same tool versions.
Corpus grows with every scan run through this instrument — 2,020 repositories today, more by the time you read this.