Brand Gravity Wellsworking title

Streaming, 2021-06 → 2026-06. The badge reads Real data · GDELT 7/12 brands because that is exactly what it is: 7 of 12 brands carry full-window GDELT affinity; the other 5 sit on neutral placeholders until their queries clear GDELT's rate limiting. The badge text is derived from the export's own coverage metadata — it cannot claim more than the data carries.

The model

• Mass M = weighted geometric mean of salience, affinity, cohesion, autonomy (0.35 / 0.25 / 0.20 / 0.20). Salience comes from Wikipedia pageview share; affinity from GDELT tone.
• Field Φ(x,y) = −Σ G·M_i/√(|x−x_i|² + ε_i²) — a softened Newtonian potential over the positioning plane. The rendered surface is cross-checked against the engine to 10⁻⁶.
• Γ = M/L (mass-to-light): attention mass vs. usage luminosity — the dark-matter detector for brands that loom larger than they are used, and the inverse.

Disclosed placeholders in this category: cohesion (M_c) and luminosity (L) have no free streaming data source yet, so both are fixed placeholders and Γ falls back to M; autonomy (M_u) currently only distinguishes covered from uncovered brands. Every placeholder is flagged per-state in the export and labeled in the tooltips — nothing placeholder is presented as measured.

Validation, honestly

Does computed mass track Nielsen Gauge share of TV time? Spearman correlations on the matched brand-months (high/medium-confidence Gauge rows), from the current build:

The pooled figure alone would overstate the evidence — its 44 rows come from 9 brands, so it is never cited without the brand-level line. What this is: evidence toward a real signal, improving as coverage grows. What it is not: proven external validation.

Data

All open sources: Wikimedia pageviews (12/12 brands), GDELT tone and volume (7/12 full-window; rate-limit-blocked queries retry weekly), a Nielsen Gauge series reconstructed from Nielsen's public releases with per-row confidence tiers (low-confidence rows excluded from validation by default), and Wikidata for roster lineage. Merger dates that are announcements rather than closings are labeled as such.