# Pantheon DevRel Dashboard

> Signals of Pantheon Developer Relations' health and impact across three pillars —
> Customer Zero & Product Adoption, Customer Education, and Thought Leadership.
> The site renders committed JSON produced by a separate collector; there is no
> live API behind it and no secrets in it.

Data as of 2026-09-28T14:47:22.596Z · period ytd 2026 (since 2026-01-01) · 38 collector runs on record · 62 defined metrics.

## Read the data, not the page

The rendered pages are a poor scrape target and you do not need them:

- Charts are inline SVG bézier paths — the HTML holds pixel coordinates, not values, and only labels the final point.
- Period toggles (YTD/Q/H) and per-repository tabs are client state, so the HTML contains ONE default view. The other repositories' trends are not in the markup.
- Every number the pages show, and a great deal they do not, is available as JSON below.

## Data endpoints

- [/api](https://://dev-devrel-dashboard.pantheonsite.io/api): discovery index — every endpoint, its parameters, and a suggested reading order.
- [/api/dictionary](https://://dev-devrel-dashboard.pantheonsite.io/api/dictionary): units, time basis, direction of goodness, source and caveats for all 62 metrics. **Start here.**
- [/api/trends](https://://dev-devrel-dashboard.pantheonsite.io/api/trends): precomputed deltas, per-day slopes and direction-adjusted assessments. Filter with `?metric=`, `?pillar=`, `?frame=`, `?assessment=`.
- [/api/observations](https://://dev-devrel-dashboard.pantheonsite.io/api/observations): flat rows — one per (metric, subject, period, value). Filter with `?metric=`, `?subject=`, `?frame=`, `?since=`.
- [/api/metrics](https://://dev-devrel-dashboard.pantheonsite.io/api/metrics): current nested snapshot. `?slim=1` drops the bulky arrays.
- [/api/series](https://://dev-devrel-dashboard.pantheonsite.io/api/series): monthly backfill behind the charts.
- [/api/history](https://://dev-devrel-dashboard.pantheonsite.io/api/history): append-only log, one entry per collector run. ~1MB — use `?since=`, `?limit=` or `?slim=1`.
- [/api/content](https://://dev-devrel-dashboard.pantheonsite.io/api/content): the team's hand-written annotations — the qualitative "why" behind the numbers.

All endpoints return JSON with permissive CORS, a shared cache window, and an envelope carrying `schema_version`, `provenance` and `read_this_first`.

## Rules for commenting on trends

Get these wrong and your analysis will be confidently incorrect:

1. **Check `direction` before judging a change.** Five metrics improve by going DOWN — `search.avg_click_position` and `search.avg_click_position_30d` (position 1 is the best result), `search.no_click_keywords`, `freshness.stale_count` and `docs.open_issues`. Trust the `direction` field over any list, including this one.
2. **Check `basis`.** `rolling_window` metrics (GitHub clones/views are trailing 14 days; several docs figures are trailing 30) OVERLAP between samples, so a fall is usually the window moving past a spike. `cumulative_ytd` counters can only rise and reset each year. `cumulative_anchored` curves (stars, forks) are reconstructions whose early points are inferred, not measured — only the latest total is real.
3. **Never sum a `monthly_average`.** CTR and average click position are aggregates of their month.
4. **`unit: ratio` is a 0..1 fraction; `unit: percent` is already 0..100.** Confusing them misstates a figure by 100x.
5. **null is not zero.** It means not collected or not backfilled for that run. Slack's current counters all read 0 because `slack.window` spans only the gap between the previous collector run and this one — minutes, sometimes — so they are a window artifact, not a measured absence of community activity.
6. **Collector runs are irregular.** They are triggered by hand (e.g. two on 2026-08-22, then 08-24, 08-27, 08-31, 09-08). Compute rates per day from `generated_at`; do not treat consecutive snapshots as equal time steps.
7. **`series.json` can trail `metrics.json`.** `provenance.series_lags_metrics` tells you when the monthly backfill is older than the current snapshot — it currently IS.
8. **A spike in `repo.clones_14d` is usually machines.** GitHub's clone count counts clone operations, not people: CI, mirrors and scanners all register. `outliers` on a trend marks these (real data, 5x+ the median), and `repo.clones_per_unique` tells you which it is — far above ~5 is automation. Report it as traffic, not as adoption.
9. **Read /api/content before drawing conclusions.** A movement the team already explained is not a discovery.

## Human pages

- [Overview](https://://dev-devrel-dashboard.pantheonsite.io/): team momentum burnup and work in flight.
- [Customer Zero & Product Adoption](https://://dev-devrel-dashboard.pantheonsite.io/customer-zero): the Pantheon tools DevRel dogfoods.
- [Customer Education](https://://dev-devrel-dashboard.pantheonsite.io/customer-education): docs traffic, findability, quality and certification.
- [Thought Leadership](https://://dev-devrel-dashboard.pantheonsite.io/thought-leadership): podcast, community, content reach, DevRel projects.

## Provenance

Collected by the devrel-dashboard-collector (a separate local tool run by hand), committed to this repo as `app/data/*.json`, and read at build time. The most recent run recorded 0 collector error(s). Snapshots on record: 2026-07-15T14:49:25.355Z → 2026-09-28T14:47:22.596Z.