Scoring Methodology

Gender Axis (0-100)

The Gender Axis scores content from 0 (strongly male-targeted) to 100 (strongly female-targeted). Five weighted factors determine the raw score:

• Genre Seed (30%) — Each genre has a baseline gender score derived from audience research. Action = 25 (male-leaning), Romance = 75 (female-leaning), Drama = 50 (balanced).
• Cast Gender Ratio (20%) — Proportion of male vs female cast members, weighted by billing position (leads count more than background).
• Lead Demographics (25%) — Gender of the protagonist(s) and primary antagonist(s). A female-led action film pulls the score toward female.
• Content Signals (15%) — Romance centrality, violence level, emotional tone, and thematic focus. High romance pulls female; high combat pulls male.
• Marketing Positioning (10%) — Trailer tone, poster imagery, tagline language, and time slot placement as indicators of intended audience.

A 1.2x stretch is applied to push scores away from the center, improving discrimination between titles that would otherwise cluster around 50.

Age Axis (0-100)

The Age Axis scores content from 0 (youth-oriented, ages 13-34) to 100 (mature-audience, ages 35+). Five weighted factors:

• Genre Seed (30%) — Animated content seeds young; political dramas seed mature.
• Maturity Rating (20%) — MPAA/TV ratings mapped to an age score. G/PG = 15, PG-13 = 35, R = 65, TV-MA = 75.
• Content Complexity (20%) — Pacing, dialogue density, narrative structure, and thematic depth. Slower, denser content scores older.
• Platform (15%) — Distribution channel age demographics. YouTube = 25, Disney+ = 30, HBO = 65, PBS = 75.
• Violence-Age Profile (15%) — Genre-aware: fantasy violence in animation seeds young, realistic violence in crime drama seeds old.

A 1.2x stretch is applied, matching the gender axis treatment.

Quadrant Scores

The four quadrant scores (Young Men, Young Women, Older Men, Older Women) are derived from the gender and age axis scores using a 1.8x stretch formula:

• Young Men = stretch(avg(100 - gender, 100 - age))
• Young Women = stretch(avg(gender, 100 - age))
• Older Men = stretch(avg(100 - gender, age))
• Older Women = stretch(avg(gender, age))

The stretch formula: clamp(50 + (raw - 50) × 1.8, 0, 100) ensures quadrant scores span the full 0-100 range rather than clustering in the 30-70 band.

Racial Audience Scores

Each title receives audience scores for 9 racial/ethnic categories (White, Black, Hispanic, Asian, South Asian, Middle Eastern, Indigenous, Mixed, Other). These use a sigmoid deviation model:

• Start with a baseline score derived from the production origin (e.g., US productions have different racial demographics than Korean productions).
• Deviate based on cast representation relative to that baseline using a sigmoid curve.
• Stronger deviations (Black-led film from a majority-White production) result in higher scores for that demographic.

Adaptation Faithfulness (0-100)

For titles adapted from source material (books, comics, games, etc.), faithfulness is scored based on:

• Plot fidelity to source material
• Character accuracy (appearance, personality, background)
• Thematic preservation
• Narrative structure alignment

Non-adaptation titles (originals) receive NULL for faithfulness — they are not scored on this dimension. Documentaries and unscripted content also receive NULL.

Advocacy Classification

The advocacy score (0-100) uses a weighted-sum formula to classify content along a purpose spectrum:

• LGBTQ Representation (20%) — Centrality and treatment of LGBTQ characters and themes.
• Disability Representation (5%) — Presence and treatment of disabled characters.
• Demographic Swaps (25%) — Race and gender swaps from source material (adaptations only).
• Advocacy Organization Involvement (10%) — Participation of advocacy groups (GLAAD, Color of Change, etc.) in production.
• Creator Statements (15%) — Explicit statements from creators about representation goals.
• Source Faithfulness (25%) — Inverse relationship: unfaithful adaptations that add representation signals score higher.

The numeric score maps to four tiers:

• Entertainment (0-20): Content made primarily to entertain with no significant advocacy signals.
• Mixed (21-40): Some representation present but not central to the production.
• Message-Leaning (41-65): Representation is intentional and noticeable, influencing story choices.
• Advocacy-Driven (66-100): Representation is a primary goal of the production, with multiple strong signals.

Audience Breadth

Each title is classified by how broadly it appeals across demographics:

• Universal — All four quadrant scores above 60 with less than 20-point spread.
• Mass Market — All scores above 45 with less than 30-point spread.
• Mainstream — Average score at or above 50 with less than 40-point spread.
• Targeted — High variance: some quadrants above 60, others below 40.
• Niche — Everything else (typically one very high quadrant with others low).

Data Confidence & Sampling

Every title carries a data-confidence value from 0 to 1 indicating how complete and reliable its underlying inputs are. Confidence rises with the number of cast members researched, the share of cast with confirmed demographics, verified genre classifications, and a populated source-material record for adaptations. Titles below 0.50 confidence are flagged as preliminary, and titles below the 0.30 indexing floor are excluded from the public sitemap until their data improves.

A note on sampling: scores describe a title's production as recorded in our data, not a survey of who actually watched it. We do not field audience polls or buy viewership panels. “Audience” here means the audience the production decisions point toward — the inferred target — which is a different and more checkable thing than measured viewership. Cast-demographic factors weight billed and leading roles more heavily than background credits, because lead casting is the strongest observable signal of intended audience.

A title moves to verifiedonly after its confidence clears the verification gate and its core scores have been accuracy-checked. The date of that check is published on each title page as a “last verified” stamp.

Data Sources

Scores combine licensed metadata with our own original research. We cite our sources so you can check them:

• TMDB — title metadata, cast and crew lists, genres, ratings, posters, and synopses for films and television, retrieved through the TMDB API. WhatSphere is independent and not affiliated with TMDB.
• IGDB, Jikan/MyAnimeList, AniList — supplementary metadata for video games and anime not well covered by TMDB.
• YouTube — channel and subscriber data for the channels we cover.
• Manual editorial research — casting changes, race/gender swaps, adaptation faithfulness, and audience-void judgments, recorded by our team from primary and reputable secondary sources.
• Creator and advocacy-organization statements — public, on-the-record statements of intent (interviews, press, organizational announcements) used as advocacy-classification inputs, attributed to their source on the relevant title and insight pages.

All scoring is original analysis derived from these inputs. We do not resell or redistribute source data.

Limitations & Known Biases

No scoring system is neutral by accident, and ours has limits we would rather state plainly than paper over:

• Genre seeds encode assumptions. Each genre starts from a baseline gender and age score derived from audience research. Those baselines are defensible but not objective truth — they bake in generalizations that will be wrong for individual titles that defy their genre.
• Demographic inference is not self-identification. Cast race and gender are coded from public records and appearance-based sources. These can be incomplete or mistaken, and they are not a substitute for how a person identifies. Roughly nine in ten cast records still lack researched race data, which is the single largest gap in our coverage today.
• Origin baselines are coarse. Racial-audience scores compare a cast against a production-origin baseline (e.g. a U.S. demographic mix). Countries are not monoliths, and the baseline is an approximation.
• Advocacy classification is interpretive. Judging whether representation is a “primary goal” of a production involves editorial judgment, even though we anchor it to weighted, observable signals. Reasonable people will disagree at the margins.
• Western and English-language coverage is deeper. Our catalog and research are stronger for titles with rich English-language metadata, so confidence skews lower for international and niche content.
• The data is a snapshot. Scores reflect the data on the date they were last verified. Casting reveals, recuts, and new research can change a score after publication.

We surface confidence levels precisely so these limits are visible rather than hidden, and the corrections process exists to fix the cases we get wrong.

Methodology Changelog

We version the methodology and record substantive changes so a score's history is transparent. Current version: v1.2, last reviewed May 31, 2026.

• v1.2 (May 31, 2026) — Published sourcing, limitations and bias disclosures; added the versioned changelog and confidence/sampling explanation; introduced "last verified" stamps tied to the verification pipeline.
• v1.1 (March 15, 2026) — Added racial-audience sigmoid deviation model and the four-tier advocacy classification; introduced data-confidence scoring and the preliminary flag below 0.50 confidence.
• v1.0 (February 20, 2026) — Initial scoring methodology: gender and age axes from weighted factors, four quadrant scores, and adaptation faithfulness.