Methodology overview

How the Caratlytics Score works

The Caratlytics Score is a 0 to 100 composite computed for certified diamonds observed on the retail market. It answers four questions: how good is the stone, how good is the price, how trustworthy is the data, and what does the market say. Each question is a pillar with its own published sub-score.

Composition

The overall score is a weighted composite. When a sub-score cannot be computed, its weight is redistributed across the available sub-scores rather than counted as zero.

Overall = Quality x 0.35 + Value x 0.30 + Certification x 0.20 + Market x 0.15

The four pillars

Gates and safeguards

Data coverage gate

A quality score is only issued when at least 60 percent of the weighted gemological evidence is present. Below that, the stone receives no quality score and is excluded from ranked results rather than scored on guesswork.

Missing data scores zero

Within the quality pillar, absent fields contribute nothing. A stone described only by its 4Cs cannot score above the low 60s, because polish, symmetry, proportions and fluorescence are unverified. Completeness is rewarded; vagueness is not.

Certification integrity gate

A stone that claims a laboratory certificate but is missing color or clarity data fails the integrity gate and its certification score is capped at 30. A certificate claim that cannot be cross-checked is not rewarded.

Price plausibility guard

Listings from vendor feeds whose prices fall below a plausibility floor of 300 USD per carat are treated as pricing artifacts. Their value score is neutralised at 50 so an implausible price can never rank as a bargain.

The evidence base

Scores are computed from the Carat Hunter dataset: live listings collected continuously from more than one hundred diamond retailers, matched across retailers by laboratory and certificate number, with full price history retained for every listing.

Category baselines (the statistical reference for the Value pillar) are recomputed daily from current market prices. Scores are recomputed when underlying data changes, typically within an hour.

Versioning

The methodology is versioned. The current version is 1.0, effective 2026-05-09. Any change to a weight, a table or a gate appears in the changelog with its rationale and effective date. Historical scores are not silently rewritten.