Methodology

Measurement of position — not advisory, not opinion.

Ragnom measures how decisions are formed inside AI systems. The methodology is designed to capture observable decision behavior — not performance, not market perception.

Controlled environments Repeatable runs +86k decisions / year 20 sectors
See the design Core dimensions
Overview

What the system isolates.

All outputs are derived from controlled, repeatable decision environments. No subjective weighting. No interpretation layer.

What Ragnom measures
  • How AI models surface options
  • How AI arbitrates between alternatives
  • How outcomes shift under constraint
What Ragnom does not measure
  • Product quality or user satisfaction
  • Financial performance or market share
  • Opinion, intent, or interpretation
Experimental Design

A structured framework for every sector.

Each sector is evaluated through a controlled decision framework separating visibility from selection, and avoiding the bias of single-mode evaluation.

01

Decision scenarios

Multiple scenarios representing real-world selection contexts are used per sector. Each scenario mirrors how a user or system would actually frame the decision.

02

Parallel evaluation

All scenarios run across several AI providers and models simultaneously. This removes single-model bias and ensures the signal reflects structural position, not a provider quirk.

03

Open recall

Unprompted citation environments measure which entities surface without any explicit framing. This captures raw cognitive presence inside AI decision space.

04

Constrained selection

Explicit comparison environments pressure-test selection under defined constraints — cost, speed, risk, complexity. This reveals substitution dynamics and resilience.

05

Repeatability

Multiple similar formulations to control semantic biases. Dozens of repetitions for each formulation to stabilize the results.

Measurement Principles

Strict constraints. No exceptions.

The system measures only what can be observed and reproduced. Signal integrity takes precedence over coverage.

Constraints applied
  • No metric interpreted in isolation
  • No subjective weighting applied
  • All computations follow deterministic rules
  • Outputs normalized within comparable sets
Consistency & validation
  • Results aggregated across models
  • Variance measured and monitored
  • Low-signal outputs filtered out
  • Stability required before interpretation
Core Dimensions

Three dimensions. Computed independently.

All entities are positioned across three standardized dimensions, then combined into a unified positioning space.

Dimension 01

Visibility

Frequency of appearance across decision scenarios. Measures how often an entity enters the set of options AI considers.

Cognitive presence
Dimension 02

Selection

Frequency of final choice. Measures how often the answer resolves in favor of the entity once the shortlist is formed.

Decision capture
Dimension 03

Resilience

Stability of selection under constraint. Measures what happens when cost, speed, simplicity, or risk begins to reshape the choice.

Pressure resistance
Decision Mapping

Each entity placed within a continuous space.

The decision map positions every entity on three axes simultaneously, reflecting how decisions are actually distributed — not how markets are described.

X axis Resistance to substitution
Y axis Cognitive presence
Size Selection frequency
Substitution Flows

Where decisions move when conditions change.

When constraints are introduced, the system records which entities lose selection and which absorb those decisions. These flows reveal structural dependencies within the decision system.

Signal type A

Exits

Which entities lose selection when cost, risk, or complexity is introduced as a deciding factor. Exits reveal over-reliance on favorable conditions.

Signal type B

Absorptions

Which entities capture decisions that move away from incumbents under pressure. Absorptions identify structural challengers before they register in traditional metrics.

Scope

One layer. Measured precisely.

Ragnom isolates a single layer: how AI systems decide. Nothing more. Nothing less.

In scope
  • Decision visibility
  • Selection outcomes
  • Substitution dynamics
  • Structural position across sectors
Out of scope
  • Product quality
  • Financial performance
  • User satisfaction
  • Market share
Output

Consistent outputs. Comparable across sectors.

Each run produces a standardized set of outputs. The format is identical across all sectors and entities — enabling direct comparison.

Output 01
Decision position
Entity placed on 3 axes within the decision space. Reflects structural position, not perception.
Output 02
Sector ranking
Relative standing within a competitive set based on selection frequency and resilience scores.
Output 03
Decision map
Visual representation of how all entities distribute across the decision space in a given sector.
Output 04
Substitution flows
Movement of decisions under constraint — who loses, who absorbs, and at what relative weight.
Objective

A consistent, comparable standard for how entities are selected by AI systems.

Ragnom is a measurement system, not an interpretation layer. It defines a reproducible framework for observing decision behavior in AI environments.

Request a reading Explore the framework