How it works

JAWNRA doesn't simply measure popularity. It measures musical momentum, using the signals we can actually observe across listening, discovery, playlists, social activity, geographic spread and independent editorial attention.

This deployment runs in demo mode. The rules below are the real engine; the artists and numbers it is currently running on are fictional fixtures.

Three kinds of fact

Everything on JAWNRA is one of three things, and we label which:

observedNumbers a provider reported to us: plays, identifications, playlist reach, creations, audience. We store them weekly and never overwrite them.
derivedThings JAWNRA calculated from observed data: growth rates, percentiles, the JAWNRA Score, chart positions, “Why it's moving”.
editorialHuman decisions: genre mappings, which publications count as editorial signal, featured discoveries.

The JAWNRA Score

A 0–100 score of how strongly a record is moving this week relative to everything else JAWNRA tracked that week. It combines seven categories:

20%Streaming momentumWeek-over-week and short-term change in listening, not total size.

20%Discovery intentIdentification and search behaviour (e.g. Shazam) — people actively trying to find out what they just heard.

15%Social momentumPermitted short-form, video and social usage signals.

15%Playlist momentumNew playlist additions and change in playlist reach.

10%Geographic spreadAppearing in new countries and cities.

10%Editorial signalMeaningful independent editorial attention from ingested, permitted sources.

10%Acceleration / noveltyRate of change of the rate of change, newness, and breakout behaviour relative to existing scale.

Rate of change, not raw size

Growth is measured on a log scale. A record going from 5,000 to 30,000 plays earns far more momentum credit than one going from 25,000,000 to 25,300,000, because the first is a six-fold change and the second is a rounding error.

Saturation correction

Above roughly 316K weekly plays, positive growth credit is progressively dampened (never below 45%), so the biggest records don't win on scale alone.

Noise floors

Growth from tiny bases is floored: below 500 plays or 25 identifications, the previous week is treated as that floor. Three plays becoming forty is not a breakout. Records with no previous week are credited as if they had half their current level, a fixed 2× rather than infinite growth.

Percentiles and dynamic reweighting

Each category is converted to a percentile across the week's field. If a category has no data for a record (for example, no permitted social data), it is left out and the remaining weights are rescaled. We never invent a missing value. A record needs at least 45% of the total weight covered by real data to be scored at all.

The JAWNRA 50

The flagship blends momentum (70%) with reach (30%, a log-scaled percentile of audience size). Momentum is smoothed with the previous week (65/35) so a single noisy week can't throw a record in or out. Major artists are not excluded: they chart when they are moving. An artist can hold at most three places.

Discovery charts

Here artist scale is part of eligibility, by design. “Audience” means the most reliable artist-level audience metric available that week. Artists without a reliable audience metric are not eligible for audience-banded charts.

20 Artists You Haven't Heard Yet: audience under 250K, no more than 2 previous weeks in the JAWNRA 50 top 30, and at least +20% growth (or a new release).
Under 10K / Under 100K / Under 1M: audience bands of 0–10K, 10K–100K and 100K–1M, ranked by momentum.
Breaking Artists: audience under 5M with at least 3 categories above the 70th percentile. Cross-platform momentum, not one lucky signal.
Rising 20: acceleration, listening and discovery growth, from a base of at least 2,000 plays.
New This Week: released within 14 days. Sleepers: released 26+ weeks ago and accelerating. Crossover: entered at least 2 new markets. Leftfield: high curiosity relative to size, genre-crossing, editorially noticed. Biggest Movers: combined upward movement across every JAWNRA chart.

Countries and cities

Country charts use only signal observed in that country, blending local momentum, local reach and over-index (how much of a record's worldwide activity comes from that market). City charts are identification-led: they rely on sources that genuinely report city-level discovery, such as licensed Shazam city data. We never take a national number and split it into cities by population.

Thresholds

A chart only publishes when enough records have valid signal: 20 for countries, 12 for cities, 10 for genres (configurable per market). When the threshold isn't met, the chart page says so. We would rather publish nothing than pad a chart.

The chart week

Charts publish every Tuesday. The tracking window is the seven days ending the Monday before. Weeks are numbered by ISO week of the publication Tuesday. Once published, a week is locked in the database: a trigger rejects any edit, so the archive is exactly what we said at the time.

What we never do

Claim access to data we don't have. The live list of sources is on Data sources.
Show a “why it's moving” reason that isn't backed by a stored signal.
Scrape sources whose terms prohibit it, or use Spotify content to train models.
Let the AI Scout invent artists, numbers or rankings. It can only return rows from the JAWNRA database.
Accept payment for chart placement.