Skip to documentation
Urdais
Documentation menu

Methodology Methodology Overview

On this page

Urdais Methodology

Purpose

Urdais methodologies define how source information becomes a published Urdais information product. Each output will be developed individually, from its definition through its data requirements to implementation.

Methodology precedes backend implementation: methodology determines what the system must know; the backend implements the system required to know it.

This document establishes the framework. It does not define a finished methodology for any individual output.

Methodology Principles

  • Traceability: connect published outputs to their inputs and transformations.
  • Reproducibility: retain enough context to explain and reproduce results where technically possible.
  • Explicit source provenance: identify original sources and how observations were obtained.
  • Versioned methodology: associate outputs with the rules effective at publication.
  • Transparent eligibility rules: document what is included, excluded, and why.
  • Historical integrity: preserve the context of past publications and document revisions.
  • Separation of data stages: distinguish raw, normalized, calculated, and published data.

Standard Output Methodology Chain

  1. Output Definition: establish what the information product represents and its intended use.
  2. Underlying Data Structure: identify the information and relationships needed to describe the output.
  3. Methodology: define how eligible information becomes the output.
  4. Source Specification: identify sources, provenance requirements, and eligibility criteria.
  5. Ingestion Contract: describe the inputs and delivery expectations required from each source.
  6. Stored Outputs: define what must be retained to support publication and historical explanation.
  7. Validation / Quality Rules: specify how completeness, consistency, and fitness for publication are assessed.
  8. Backend Implementation: implement the system required by the preceding decisions.

Output Methodology Template

Copy these headings when developing an individual output. Leave unresolved decisions explicit; the headings below do not prescribe schemas, calculations, sources, or publication schedules.

# [Output name] Methodology

## Output Identity

## Output Definition

## Published Surface

## Underlying Data Structure

## Methodology

## Source Specification

## Ingestion Contract

## Stored Outputs

## Validation / Quality Rules

## Lineage and Reproducibility

## Publication and Revision Behavior

## Methodology Version

## Version History

Methodology Categories

The following categories are reserved for future output-specific methodologies. None is defined by this framework.

  • Indices: to be developed one output at a time. The first proposed index methodology is UGAI, the Urdais Global AI Index, which consumes the shared AI Equity Universe. UAVI, the Urdais AI Volatility Index, is a proposed sibling output that consumes the same universe and measures 30-day option-implied volatility across its option-eligible members. UBWI, the Urdais Bitcoin Wealth Index, is a separate proposed index that shares no universe with either: it measures Bitcoin's market capitalization as a percentage of Total Global Wealth, and its value is a percentage rather than an index level. UTVI, the Urdais Observed Token Volume Index, is a production methodology for observed model token consumption per day, approved at 1.0.0 and publishing since 16 September 2026. It shares no universe with the others: it measures the token volume one marketplace's dataset exposes, and its universe statement defers to that source rather than asserting what the source does not document. Market Share is its derived breakdown, versioned with it at 1.1.0: share of that same observed volume by model and by canonical lab, with both residuals reported rather than absorbed.
  • Compute prices: UCPI, the Urdais Compute Price Index family, is a separate shared methodology for measuring what a defined unit of market-accessible accelerated compute costs. It is a parent for per-instrument children rather than a single index, and it does not consume the AI Equity Universe.
  • Markets: to be developed one output at a time. The first is UMPI-KR DRAM, version 1 of the Urdais Memory Price Index: two monthly official-data DRAM indexes built from Korean government statistics, in index points, with a MoM change. Its predecessor UMPI-DRAM Spot — six DRAM chips priced daily in USD per chip from a commercial spot board — is deferred rather than withdrawn: the definition is complete, the rights are not, and no open source prices those instruments at all.
  • Market Analytics: to be developed one output at a time.
  • Maps: to be developed one output at a time.
  • News / Information: to be developed one output at a time.

Shared Methodologies

The Urdais AI Equity Universe proposes shared company eligibility and base weighting for future UGAI and UAVI outputs. It is a draft with unresolved empirical parameters, not a production universe or a completed index methodology.

Lineage Model

Published Output → Methodology Version → Calculated / Normalized Inputs → Raw Observations → Original Sources

Future Urdais outputs should be historically explainable and reproducible where technically possible. Each output's methodology should describe which stages apply and document any limits to reconstructing its lineage.

Versioning

Each output methodology should have a version identifier, an effective date, documented material changes, and a historical changelog. Published outputs should remain attributable to the methodology version that produced them; subsequent changes should not silently rewrite that history.