Traffic signals
Views count repository page requests reported by GitHub, while unique views are GitHub's aggregate estimate of distinct viewers for an interval. Clones count full repository clones rather than fetches; unique clones are the corresponding aggregate unique estimate.
Referrers show leading external sources, excluding search engines and GitHub itself in GitHub's traffic interface. Popular paths show which repository content generated traffic. Both lists are ranked by views and unique visitors and cover the previous 14 days.
Growth signals
Stars are a lightweight expression of interest, not proof of usage. RepoLogbook stores daily cumulative star observations and derives non-negative interval gains from adjacent successful snapshots. Forks use the same snapshot-difference approach.
Genuine subscribers use GitHub's subscribers count: people who explicitly watch repository activity. That is distinct from the legacy watchers field, which aliases stars. Optional release and npm download signals refresh only through their supported, owner-requested workflows.
- Keep traffic, adoption, and distribution signals distinct
- Show source and freshness with each optional metric
- Do not convert aggregate counts into people or customers
- Treat unavailable observations as unavailable
Compare projects without flattening them
A cross-repository report uses one metric, date range, granularity, chart mode, and scale across the selected repositories. Each repository keeps its own series while the report also shows the combined total.
Interval, cumulative, normalized, and rolling modes answer different questions. Normalized views compare relative shape, not absolute audience size. Log scales help with large magnitude differences but require special handling for zero values. The accessible table preserves exact values outside the chart.
Save, share, and interpret
Saved reports retain a query definition for reuse. Public reports publish a sanitized immutable snapshot with no tenant ID, internal repository ID, or account details. README cards are rendered ahead of time and served from Blob storage without executing a Function for each view.
A report is evidence of observed repository-level change. It is not proof that a campaign, release, or person caused that change. Use annotations and date comparisons as context, then validate important decisions with additional evidence.
Frequently asked questions
Which GitHub repository metrics belong together?
Views, unique views, clones, unique clones, referrers, popular paths, stars, forks, and genuine subscribers answer different questions. Compare related signals rather than collapsing them into one score.
Are unique visitors the same people across days?
No. GitHub provides aggregate unique estimates for each reported interval. Those counts should not be summed into identified people.
Can repositories be compared in one report?
Yes. RepoLogbook can combine selected tracked repositories under the active plan's report-size limit and show daily, weekly, or monthly series.
Are public reports live dashboards?
No. Owner-published public reports are immutable snapshots. They include their generation time and can be revoked.