Repository metrics

Ecosystem adoption

How inspect.software measures real package adoption — registry downloads and dependents across npm, PyPI, Packagist, and more. 4.5% of the overall index.

Methodology v1.13.0Updated 2026-07-13

Ecosystem adoption measures how widely a package is actually installed, using the package registries themselves — npm, PyPI, Packagist, crates.io, RubyGems, Hex, and the other supported ecosystems. Real download volume beats GitHub stars as adoption evidence: downloads measure use, stars measure attention.

  • Category: Community & Adoption (25% within category)
  • Weight in overall index: 4.5%
  • Metric key: ecosystem_adoption
  • Applies to: repositories that publish at least one package; otherwise null

How the value is computed

ComponentWeightCriteria
Monthly downloads80summed across the repository's packages, log-scaled — ~1,000,000/month saturates
Total downloads80fallback when a registry publishes no monthly figure (e.g. RubyGems); log-scaled, ~50,000,000 all-time saturates
Registry dependents20log-scaled; reported by some ecosystems only, excluded elsewhere

The download component uses the monthly figure where available and the lifetime total otherwise — whichever the registry provides, so no ecosystem is structurally disadvantaged.

Fairness rules

  • Non-publishing repositories are not penalized. Applications and infrastructure projects that ship nothing to a registry get null here, and Community & Adoption renormalizes onto its other metrics.
  • Packages must point back. A registry package only counts when its metadata verifiably references the inspected repository; mismatched packages are listed in the report but excluded from the computation.
  • Log scaling keeps the metric meaningful across the whole range — a package going from 1,000 to 10,000 monthly downloads moves as much as one going from 100,000 to 1,000,000.

Reading the result

  • Within its category, this is the strongest adoption evidence available — stronger than popularity, which is why a heavily downloaded but modestly starred package can outread a famous one.
  • Download figures vary by registry in meaning (mirrors, CI traffic); the log scale absorbs most of that noise, and the exact inputs are echoed in every report for inspection.

Improving the value

  • Publish the package to its ecosystem's registry and keep the repository link in the package metadata accurate.
  • Beyond that, downloads follow genuine use — the honest lever is the project's usefulness and discoverability, not the metric.

Related: package maintenance · supported ecosystems