Skip to content
CANONICAL TERMINOLOGY

The IEO Glossary

The canonical terminology of Interpretability Engine Optimization (IEO), organized by conceptual layer — from foundational terms to measurement, boundaries and economic language.

Organized by conceptual layer rather than alphabetically, because the terms build on each other. Every term has its own permanent anchor — cite it directly.

Version 2.2 · 59 terms · English canonical

Download the glossary (PDF) →18 pages · CC BY 4.0

Layer 1

Foundation

Interpretability

Definition: The degree to which a machine system can form a correct, usable understanding of business information from the representation available to it.

Example: A product page may clearly show a sunscreen to a human, while a machine may still fail to resolve its product type, SPF, intended use, active ingredients, or restrictions.

Boundary: In IEO, interpretability means machines interpreting the merchant. This is the reverse direction of model interpretability in XAI, where humans interpret the model.

IEO — Interpretability Engine Optimization

Definition: The practice of making a business's offers, entities, attributes, policies, and source facts machine-interpretable before downstream retrieval, recommendation, comparison, or action occurs.

Operational formulation: IEO structures, resolves, enriches, validates, and represents source business data so machine systems can correctly understand the offer without inventing missing facts.

Category status: Open discipline. IEO does not prescribe a specific vendor, model, data format, protocol, or software implementation.

Canonical framing: IEO optimizes understanding.

Canonical closing line: Semantic discovery is downstream of interpretation.

Scope of the acronym: "IEO" is used here for Interpretability Engine Optimization, the discipline proposed by NSOLVIA Research in 2026 and published under DOI 10.5281/zenodo.22104280. The acronym has other expansions in circulation; this work refers exclusively to the one defined here.

Research note: What the term refers to, where it sits relative to neighbouring practices, and what falls outside its scope are set out in Interpretability Engine Optimization: Terminology and Scope.

Commerce IEO

Definition: The application of IEO to products, catalogs, brands, offers, and commerce operations.

Scope: Commerce IEO is the domain in which the current NSOLVIA research, instruments, and empirical evidence are most mature.

Interpretability Engine

Definition: A collective term for any machine system that consumes business information and must form a usable interpretation of it, including search engines, answer engines, generative systems, recommendation systems, commerce platforms, and autonomous agents.

Clarification: IEO optimizes source data for interpretation engines. It does not optimize the engines themselves.

Commerce Interpretability

Definition: The umbrella condition describing whether machines can correctly understand the three objects required to evaluate a commerce offer: the product, the brand, and the transaction.

Canonical model: Product tells the machine what is being sold. Brand tells it who stands behind it. Transaction tells it under what rules it can be bought.

AI Commerce Readiness

Definition: The state of preparedness of a merchant's commerce data and surfaces for machine-mediated discovery, interpretation, comparison, recommendation, and action.

Clarification: Readiness is a state. IEO is a practice used to improve one of the foundational properties that readiness depends on.

Boundary: Distinct from any vendor-specific readiness score: commercial instruments (e.g., NSOLVIA's Agentic Catalog Readiness Score™) measure readiness including interpretability-related signals, but no single commercial score should be read as a universal IEO metric.

Interpretability Debt

Definition: The accumulated gap between what a business knows about its offers and what its machine-facing representations make explicit enough for machines to understand correctly.

Example: Years of adding products for human shoppers while leaving taxonomy, attributes, use cases, policies, and structured meaning unresolved can create interpretability debt.

Commercial shorthand: A catalog can accumulate interpretability debt just as a software system accumulates technical debt.

Layer 2

Objects of Interpretation

Product Interpretability

Definition: The degree to which a machine can correctly understand what is being sold, including identity, taxonomy, attributes, materials, use cases, functional intent, restrictions, and meaningful differences between related products.

Brand Interpretability

Definition: The degree to which a machine can correctly understand who stands behind an offer, including brand identity, provenance, positioning, manufacturing facts, certifications, supported claims, promises, and trust-relevant facts.

Transactional Interpretability

Definition: The degree to which a machine can correctly understand the rules under which an offer can be evaluated or purchased, including price, availability, shipping, returns, eligibility, exclusions, fulfillment conditions, and other transaction rules.

Offer Interpretation

Definition: The complete understanding a machine forms about an offer across product, brand, and transactional context.

Clarification: The unit of analysis is not merely the URL or SKU. It is what the machine can correctly understand about the complete offer.

Layer 3

Source Truth & Grounding

Source Truth

Definition: The merchant-supported facts available from the source materials from which machine-interpretable representations may be derived without introducing unsupported claims.

Merchant Truth

Definition: The complete set of facts, policies, claims, and operating rules that the merchant has supplied, published, or explicitly confirmed as authoritative.

Relationship: Source Truth is the evidence available in a given source. Merchant Truth is the broader authoritative business truth from which multiple sources may derive.

Entity Grounding

Definition: Connecting a machine-resolved entity to sufficient evidence to establish what real product, brand, organization, place, or offer the data refers to.

Brand Entity Resolution

Definition: Resolving scattered references, names, descriptions, claims, and identifiers into a consistent representation of one brand entity.

Brand Provenance

Definition: Verified information describing the origin of a brand, product, manufacturing process, material, claim, or other trust-relevant fact.

Brand Claim Grounding

Definition: Connecting a brand claim to merchant-supported or otherwise verified evidence before representing that claim as fact.

Example: “Founded by doctors” is not treated as a verified brand fact unless the authoritative source supports it.

Brand Policy Resolution

Definition: Converting merchant policies such as shipping, returns, warranties, exclusions, contact methods, and fulfillment rules into explicit, internally consistent rules that machines can interpret.

Brand Promise Resolution

Definition: Distinguishing supported brand promises from aspirational or promotional language and representing only the claims that can be sustained by available evidence.

Verified, Never Invented

Definition: Information may be normalized, resolved, or made explicit, but unsupported facts must not be introduced as if they were verified merchant truth.

Type: IEO principle.

Honest Empty

Definition: A field intentionally left unresolved because the available evidence is insufficient to support a reliable value.

Example: If material composition cannot be established from merchant-supported evidence, IEO leaves the material unresolved rather than guessing.

The Mirror Principle

Definition: Human-facing and machine-facing representations should communicate the same underlying facts because both derive from the same authoritative truth.

Purpose: Prevents machine-facing representations from becoming a hidden or contradictory version of the merchant's public offer.

Layer 4

Semantic Resolution & Catalog Operations

IEO Pipeline

Definition: The canonical typical sequence used to transform source business information into a machine-interpretable representation.

Canonical pipeline: Normalization → Semantic Resolution → Enrichment → Validation/Grounding → Machine-Interpretable Representation

Implementation rule: This is a typical IEO pipeline, not a required software architecture. It may be executed by humans, deterministic rules, AI systems, semantic engines, or any combination. IEO depends on the measurable interpretability outcome, not on a specific implementation.

Semantic Resolvability

Definition: The degree to which the meaning required for machine interpretation can be resolved from available evidence without unsupported inference.

Semantic Resolution

Definition: The operation of converting ambiguous, implicit, inconsistent, or fragmented business information into an explicit and sufficiently supported meaning.

Category Resolvability

Definition: The degree to which an offer can be reliably mapped to the most specific supported category within the target taxonomy.

Leaf-level rule: Resolve to the deepest supported category only when the evidence justifies it. Leaf-level resolution is part of Category Resolvability, not a separate discipline term.

Attribute Completeness

Definition: The degree to which the attributes materially relevant to interpreting an offer are explicitly represented.

Clarification: Completeness does not mean indiscriminately filling every available field. Unsupported fields remain Honest Empty.

Attribute Normalization

Definition: Converting equivalent attribute values into a consistent canonical representation without changing their underlying meaning.

Example: SPF45, SPF 45, and 45 SPF may normalize to the same canonical value.

Variant Resolution

Definition: Correctly distinguishing and relating purchasable variants of an offer, including size, color, material, configuration, pack, or other variant-defining properties.

Offer Integrity

Definition: The condition in which the product identity, attributes, brand facts, transaction rules, and supporting evidence remain internally consistent across the machine-interpretable representation.

Semantic Ambiguity

Definition: A condition in which the available evidence supports more than one plausible interpretation and no single interpretation can be safely resolved.

Catalog Ambiguity

Definition: Semantic ambiguity occurring at product or catalog level, including unclear taxonomy, overlapping product identity, unresolved variants, inconsistent attributes, or conflicting source descriptions.

Usage: Product/catalog-specific term. Use Semantic Ambiguity when the concept applies more broadly to product, brand, or transaction data.

Normalization

Definition: The process of making inconsistent representations consistent without adding new substantive meaning.

Enrichment

Definition: The process of making useful meaning explicit by adding supported structured information that was previously implicit, fragmented, or absent from the representation.

Boundary: Enrichment is an IEO operation when it improves machine understanding and remains grounded in supported truth.

Validation / Grounding

Definition: The operation of checking that a resolved or enriched representation is supported by authoritative evidence and does not introduce contradictions or unsupported facts.

Layer 5

Machine-Interpretable Artifacts

Machine-Interpretable Representation

Definition: A representation in which enough supported meaning has been made explicit for a machine to reliably interpret the relevant object or offer.

Canonical Offer Representation

Definition: A consistent, normalized representation of an offer that preserves its authoritative identity, attributes, brand context, and transaction rules across downstream destinations.

Semantic Product Record

Definition: The product's verified meaning made explicit.

Typical contents: Resolved identity and taxonomy, normalized attributes, functional intent, use cases, purchase signals, safety information, and other supported product meaning.

Semantic Brand Record

Definition: The brand's verified identity and supported meaning made explicit.

Typical contents: Brand identity, provenance, positioning, manufacturing facts, supported claims, certifications, promises, and trust-relevant facts.

Semantic Transaction Record

Definition: The verified transaction rules governing an offer made explicit for machine interpretation.

Typical contents: Price, availability, shipping, returns, eligibility, exclusions, fulfillment conditions, and related commerce rules.

Source Record

Definition: The merchant information as it arrives before IEO processing or remediation.

Ingested Record

Definition: The representation produced from the available source after the system has resolved only what the evidence safely supports.

Clarification: An ingested record is not automatically the final or gold representation.

Gold Record

Definition: The verified, curated representation after resolution, remediation, and — where source evidence was insufficient — authorized human confirmation.

Layer 6

Measurement & Testing

Resolution Tasks

Definition: Observable tasks used to test whether a machine can correctly resolve the meaning required to interpret an offer.

Initial framework: 1. Identity resolution · 2. Taxonomic resolution · 3. Attribute resolution · 4. Intent resolution · 5. Use-case resolution · 6. Trust / safety resolution · 7. Retrieval-language resolution

Status: Initial observable framework, not a closed taxonomy.

Interpretability Baseline

Definition: The observed interpretability state of an offer or business representation before remediation.

Projected Interpretability Lift

Definition: The estimated improvement in interpretability represented by a remediated or enriched version before that improvement has been fully deployed and re-measured in production.

Observed Interpretability Lift

Definition: The measured difference in interpretability after remediation has actually been deployed and the representation has been re-evaluated.

Interpretation Failure Cost

Definition: The downstream operational or commercial cost that may arise when a machine interprets an offer incorrectly or incompletely.

Possible effects: Missed eligibility, incorrect matching, recommendation errors, avoidable support work, avoidable returns, cancellations, or additional manual remediation.

Rule: These effects must be measured where claimed; they are not automatic consequences of every interpretation failure.

IEO Counterfactual Product Test

Definition: A controlled experiment that compares the same product represented in a fully interpreted form and a deliberately degraded form while holding material variables constant.

Purpose: Tests whether representation quality materially affects machine understanding.

Typical measures: Attribute extraction accuracy, category resolution, constraint satisfaction, differentiation between similar offers, hallucination rate, and task completion.

Layer 7

Boundaries & Neighboring Disciplines

The IEO Ladder

Definition: A functional map showing how IEO relates to neighboring optimization disciplines and the downstream agentic action layer.

Canonical ladder: IEO → UNDERSTAND · SEO → FIND · AEO → ANSWER · GEO → CITE / SURFACE · Agentic → ACT

Clarification: These are functional centers of gravity, not impermeable borders. IEO addresses machine understanding of the source representation; neighboring disciplines and agentic systems may overlap with interpretability operations, but their primary outcomes differ.

Boundary Test

Definition: The test used to determine whether an intervention belongs to IEO.

Canonical question: Did the intervention improve the machine's understanding of the source — or did it only improve how the source is formatted, distributed, discovered, or surfaced?

Practical test: If the same information were delivered through a different channel tomorrow, would the improvement in understanding remain?

Visibility

Definition: The downstream condition in which a business, product, or offer is surfaced, cited, ranked, recommended, or otherwise exposed by a system.

Relationship to IEO: Visibility is an outcome layer. Interpretability is a source-representation property that can influence downstream machine-mediated decisions but does not guarantee visibility.

Interpretability vs. Visibility

Definition: Interpretability asks whether the machine can correctly understand the source. Visibility asks whether the system ultimately surfaces it.

Canonical distinction: Interpretability is a condition of understanding; visibility is a downstream result.

Structured Data

Definition: A machine-readable encoding format or schema used to represent information.

IEO boundary: Structured data may carry interpretable meaning, but formatting alone is not IEO. A record may be valid JSON-LD and still remain semantically ambiguous.

AEO — Answer Engine Optimization

Functional center of gravity: ANSWER — improving the ability of answer-oriented systems to extract and formulate useful answers from available information.

Agentic Readiness

Functional center of gravity: ACT — enabling agents to correctly compare, decide, and execute actions using sufficiently interpretable and executable commerce information.

Relationship: IEO addresses the understanding layer on which reliable agentic action depends.

Layer 8

Economic Language

Economic Translation of Interpretability

Definition: The process of connecting a measured improvement in machine interpretation to observable operational, platform, or commercial outcomes.

Canonical chain: Interpretability Baseline → Interpretability Lift → Operational Effect → Platform Effect → Commercial Effect

Rule: Do not jump from interpretability improvement directly to revenue claims. Measure each bridge where possible.

IEO Glossarysits inside NSOLVIA's interpretability stack — an open discipline, the framework that implements it, and the instruments that apply it.

Keep reading