Methodology
Every number here is constructed. Here is how.
The Culture Index is a weighted composite of five measures, rebased daily on a 0–100 scale. Methodology changes are versioned and dated. Historical scores are never silently recomputed.
“Velocity”
WEIGHT 30%How fast is attention accumulating?
Inputs · Wikipedia pageview deltas, GDELT article volume by theme, RSS coverage counts, YouTube view and comment velocity.
Formula · z-score of (24h volume ÷ trailing 28-day median), clipped to ±4, mapped to 0–100.
“Spread”
WEIGHT 25%How many distinct communities carry it?
Inputs · Distinct media verticals in RSS coverage, distinct GDELT source countries and language codes, distinct YouTube channel categories.
Formula · Normalised Shannon entropy across community buckets. A topic confined to one vertical scores near zero no matter how loud it is.
“Sentiment”
WEIGHT 20%What is the valence and intensity of the conversation?
Inputs · Model classification of a sampled corpus of article ledes and comment samples, plus GDELT tone.
Formula · ((valence + 1) ÷ 2) × intensity, rescaled to 0–100. Stance — adopt, reject, mock, debate — is reported separately and never collapsed.
“Commerce Pull”
WEIGHT 15%Is attention converting to demand?
Inputs · Public app rank feeds, public retail and resale rank movement, reported earnings-call mentions, linked-ticker price and volume.
Formula · Weighted rank-delta composite, winsorised at the 5th and 95th percentiles.
“Durability”
WEIGHT 10%Is it compounding or decaying?
Inputs · 28-day attention series per entity.
Formula · Exponential decay fit, reported as half-life in days. Long half-life with rising velocity is a structural shift; short half-life is a meme.
“Movement Classification”
- EMERGINGvelocity > 70, spread < 40
- BREAKING OUTvelocity > 70, spread > 60
- MAINSTREAMspread > 75, velocity 30–70
- COMMERCIALIZINGcommerce pull rising 3 consecutive days
- FADINGhalf-life < 5d, velocity < 30
- BACKLASHmock + reject stance share > 40%
“Confidence”
HIGH
4 or 5 sub-indices available, 3+ independent sources agree on direction
MEDIUM
3 or 4 sub-indices available
LOW
3 sub-indices available and sources disagree on direction
NOT PUBLISHED
fewer than 3 sub-indices available — no composite, gap in the chart
A gap in the chart is more credible than a guess. Confidence is driven by source coverage, not by model certainty.
“Tracked Universe”
18 entities across four domains. Depth beats breadth; adding a domain is a configuration change, not a rebuild.
Fashion & Beauty
6 ENTITIES
Music & Audio
5 ENTITIES
Internet Native
4 ENTITIES
Food & Beverage
3 ENTITIES
Not yet tracked
Film / TV / Streaming · Sport & Athlete Brand · Gaming & Virtual Worlds · Wellness & Identity · Values-Based Consumer Behavior · Technology in Culture
“Demo Dataset Disclosure”
A synthetic demo dataset exists and is currently active. It generates 180 days of plausible history so the product can be evaluated before live history accumulates.
- It is watermarked at all times by a fixed ribbon and a DEMO chip on every panel.
- It lives in a single isolated module and is never written to the database.
- It can never be reached from the live data path.