How AnyCheese Ranks Cheeses
AnyCheese rankings are designed to be data-driven, repeatable, and transparent.
They do not measure taste or quality. They measure fit, definition, and interest.
Each ranking follows the same mechanical system.
Step 1: Eligibility Filtering
Every ranking begins by defining which cheese types are eligible.
Eligibility rules may include:
- Country of origin — for regional lists (e.g., French cheeses)
- Texture or style — for category lists (e.g., soft cheeses)
- Milk type — for milk-based lists (e.g., goat's milk cheeses)
- Minimum data requirements — cheeses must have sufficient information
Only cheese types that pass all applicable filters are considered for scoring.
Step 2: Five Scoring Signals
Each eligible cheese type receives a score based on five independent signals:
A. Category Purity (0-3 points)
For texture-based rankings (e.g., "soft cheeses"), this measures how precisely a cheese matches the category.
- Exact match: 3 points (e.g., "soft" matches "soft")
- No match: 0 points (e.g., "semi-soft" does not match "soft")
This is an all-or-nothing score. Partial matches are not counted to ensure category focus.
B. Data Completeness (0-3 points)
Cheese types are rewarded for having comprehensive, well-defined information.
Five data categories are checked (maximum 3 points awarded):
- Country of origin or specific regional origin
- Milk type (cow, goat, sheep, etc.)
- Texture (soft, hard, semi-soft, etc.)
- Flavor or aroma description
- Age or rind information
Each populated category earns 1 point, capped at 3 total.
C. Popularity by Page Views (0-2 points)
Measures how often people view a cheese type's detail page.
Views are bucketed into three tiers using percentiles:
- Top tier (above 67th percentile): 2 points
- Middle tier (33rd-67th percentile): 1 point
- Bottom tier (below 33rd percentile): 0 points
Percentiles are calculated within each ranking's eligible set.
D. Search Demand (0-2 points)
Measures how often users explicitly select a cheese type from autocomplete search results.
Only confirmed selections count:
- User types in search box and clicks a suggested cheese type
- Free-text searches without selection are excluded
- Partial inputs are excluded
Like views, search selections are bucketed into three tiers:
- Top tier (above 67th percentile): 2 points
- Middle tier (33rd-67th percentile): 1 point
- Bottom tier (below 33rd percentile): 0 points
E. Authority Bonus (0-1 point)
Cheese types with protected geographical status (PDO, PGI, etc.) receive a 1-point authority bonus.
This recognizes officially recognized cheese varieties.
Step 3: Weighted Scoring
Each of the five signals can have a configurable weight multiplier.
The final score is calculated as:
Total Score = (A × weight_purity) + (B × weight_completeness) + (C × weight_views) + (D × weight_search) + (E × weight_authority)
Different rankings may emphasize different signals, but the calculation method remains consistent.
Step 4: Sorting and Selection
Cheese types are sorted by total score in descending order.
Tiebreaker: If two cheese types have the same score, the one with more page views ranks higher.
The top N cheese types (typically 25) are selected for the final ranking list.
What the Rankings Are — and Aren't
They are:
- Based on real user behavior and data completeness
- Mechanically calculated using consistent rules
- Updated automatically as data changes
- Transparent and auditable
They are not:
- Taste judgments or quality ratings
- Expert opinions or editorial picks
- Sponsored or paid placements
- Static lists
Why This Approach
Cheese discovery has long relied on awards, reviews, and subjective lists.
AnyCheese takes a different approach: measuring how people actually interact with cheese information at scale.
The goal is not to decide what tastes best—but to surface what is most relevant, well-defined, and sought after.
By combining category precision, data quality, and real user interest signals, the rankings reflect both substance and demand.