# Methods & Limitations — Community Demand, Sentiment & Evidence Index (v1.0.0)

## What this is
A cross-compound index combining three independent signals per performance-enhancement
compound: **community sentiment** (from curated public voices), **popularity/demand**
(from search volume), and **evidence strength** (Enhanced Studies' gauge). Aggregate and
anonymized. Not clinical, not a safety assessment.

## Sources & method
- **Sentiment** — 1681 curated public voices across 72 compounds (expert/clinician "pro",
  forum "community", "influencer"), each stance-coded during curation as
  v=positive / x=negative-or-caution / m=mixed (mapping per `build.mjs`). We report counts
  and shares, overall and split by source type. **Only aggregate counts are published — no
  voice names and no quotes** (those appear on-site as attributed citations; an automated
  leak-check keeps them out of this dataset).
- **Demand** — total monthly search volume summed across each compound's keyword set
  (keyword-research corpus). This is the honest popularity signal; `n_voices` is NOT
  (it is harvest-capped). Coverage: 47/72 compounds have search-volume data.
- **Evidence** — the Enhanced Studies evidence gauge (study quantity × quality; see the
  Methods page). Included so demand can be read against evidence.

## Limitations (read before citing)
- **Sentiment is from curated public voices**, not a representative survey; selection and
  curation bias apply. Shares are indicative.
- **`n_voices` is capped by harvest depth** and must NOT be read as popularity — use `demand`.
- **Demand** is aggregate search volume (a proxy for public interest), not sales or usage.
- **Cross-tabs** (e.g. high-demand + weak-evidence) describe *attention vs evidence*, not
  safety or efficacy claims.
- No personal data; educational/informational only; **not medical advice**.
License: CC-BY-4.0. Never fabricated: where a signal is missing it is left null, not guessed.