The Five Domains

Every brand is scored across five cultural domains, each worth 0 to 20 points, for a total score out of 100. Two domains use real-time API data. Three use independent AI analysis.

A / 20 pts

Attention

Who is being looked up? Measured via Wikipedia pageviews over the past seven days. The most democratic indicator of real-time public curiosity.

Source: Wikipedia Pageviews API
B / 20 pts

Conversation

Who is being discussed? Reddit posts, comments, and upvotes from the past week. What real people choose to talk about, unprompted.

Source: Reddit API
C / 20 pts

Creation

Who is shaping culture? Creative influence, design impact, cultural reference points. The brands that inspire what gets made next.

Source: GPT-4o-mini (independent AI call)
D / 20 pts

Desire

Who do people want? Consumer demand, aspirational value, cult followings. The difference between recognition and genuine wanting.

Source: GPT-4o-mini (independent AI call)
E / 20 pts

Zeitgeist

Who is in the conversation now? Current cultural moment. The brands generating headlines, memes, and discourse right now in 2026.

Source: GPT-4o-mini (independent AI call)
Supplementary

Market

Financial performance proxy. Market cap and 30-day stock price change for publicly traded brands. Context, not score.

Source: yfinance (does not affect score)

The Hybrid Model

The Relevance Index uses a hybrid scoring model that deliberately separates quantitative signals from qualitative analysis. This prevents the common problem of AI models anchoring on a single data point.

Step 1

Data Collection

Wikipedia pageviews and Reddit mentions are pulled via APIs for each brand.

Step 2

AI Analysis

Three independent GPT-4o-mini calls score Creation, Desire, and Zeitgeist separately to prevent score anchoring.

Step 3

Score Assembly

All five domain scores are combined. Market data is appended as supplementary context.

Step 4

Deploy

Scores, insights, brand pages, and OG images are generated and deployed automatically.

The key design decision: each AI dimension gets its own focused API call. Rather than asking a single prompt to score all three qualitative domains at once (which leads to score clustering), Creation, Desire, and Zeitgeist are each scored in isolation. The AI only sees the brand name and the single dimension it needs to evaluate.

Why This Approach

Most "brand ranking" lists rely on either pure data (social mentions, web traffic) or pure opinion (expert panels, surveys). Both have blind spots. Data misses nuance. Opinion misses scale.

The hybrid model attempts to get the best of both: quantitative signals for attention and conversation (things that can be measured objectively) combined with AI qualitative analysis for creation, desire, and zeitgeist (things that require cultural judgement).

Wikipedia pageviews are used instead of social media follower counts because they represent genuine curiosity rather than passive following. Reddit is used because discussions there tend to be longer-form and more substantive than most social platforms.

Score Tiers

75 - 100 Culturally Dominant. The brands everyone is talking about and that actively shape the cultural conversation.
55 - 74 High Relevance. Strong cultural presence with meaningful influence across multiple domains.
35 - 54 Moderate Signal. Recognisable brands with pockets of cultural activity but not yet dominant.
15 - 34 Low Signal. Present but not generating significant cultural conversation or influence.
0 - 14 Off Radar. Minimal cultural footprint in the current cycle.

Update Frequency

Schedule
Weekly

Every Wednesday at 3am UTC via automated GitHub Actions pipeline.

Brands Scored
1,203

Across 9 categories: Tech, Fashion, Food, Culture, Sports, Media, Auto, Finance, People.

Stock Data
~260

Publicly traded brands enriched with market cap and 30-day price change.

AI Calls Per Run
3,600+

Three independent GPT-4o-mini calls per brand (Creation, Desire, Zeitgeist).

Caching and Freshness

To balance cost and freshness, the pipeline uses a multi-layer caching strategy:

API data (Wikipedia, Reddit) is cached for 7 days. This means Attention and Conversation scores reflect the previous week's activity.

AI scores (Creation, Desire, Zeitgeist) are cached for 30 days. Cultural influence and desire shift more slowly than attention, so a longer cache window is appropriate.

Financial data (stock prices, market caps) is cached for 7 days.

Limitations

No scoring model is perfect. The Relevance Index is an experiment in measuring something inherently subjective. Some known limitations:

English-language bias. Wikipedia and Reddit data skew heavily toward English-speaking audiences. Brands with strong cultural relevance in non-English markets may be underscored.

AI model limitations. GPT-4o-mini has a knowledge cutoff and may not capture very recent cultural shifts in its qualitative scoring.

Category weighting. All five domains are weighted equally (20 points each). A brand that dominates attention but has low creative influence scores the same as the reverse. This is a deliberate design choice: cultural relevance requires breadth.

Built By

The Relevance Index is a project by Mike Litman. The entire pipeline, from data collection to scoring to deployment, is automated and runs weekly without manual intervention.

Explore the Index