Glossary of Terms¶

Accessibility Persona¶

A profile that includes a specific disability or access need to test barriers in content, interaction, or service.

Example: A keyboard-only user attempts signup without a mouse and finds focus trapped in the plan selector.

Accessibility Score¶

A rating of how readily people with varied abilities can perceive, understand, navigate, and act on an asset.

See also: Persuasiveness Score, Evaluation Baseline

Adversarial Persona¶

A profile intentionally inclined to challenge, misuse, or strongly resist an offer so weaknesses can be stress-tested.

Example: A highly suspicious buyer tests whether vague security claims survive persistent questioning.

See also: Edge-Case Persona, International Persona

Adversarial Questions¶

Deliberately challenging questions used to test claims, expose weaknesses, or examine behavior under pressure.

Example: The skeptic asks what evidence supports “best in class” and what would disprove the claim.

Advertisement Evaluation¶

Assessment of an advertisement’s stopping power, message comprehension, relevance, credibility, brand linkage, and intended response.

Agent Autonomy¶

The degree to which an agent may choose actions or next steps without requesting human approval.

Agent Boundaries¶

Explicit limits on an agent’s allowed actions, data access, claims, decisions, and escalation authority.

Agent Context¶

The task-relevant facts, source material, and prior state made available to an agent for one assignment.

Agent Coordination¶

The rules and mechanisms that sequence agent work, route inputs, share approved results, and prevent duplicated or conflicting actions.

Agent Debate¶

A structured exchange in which AI participants argue different interpretations to test evidence and reveal disagreement.

Agent Goal¶

The outcome an agent is expected to pursue when completing its assigned task.

Agent Handoff¶

The controlled transfer of work, context, evidence, and responsibility from one agent or workflow stage to another.

See also: Agent Workflow, Agent Coordination

Agent Independence¶

The degree to which an agent reaches its initial judgment without exposure to other agents’ opinions.

Agent Instructions¶

The task-specific rules an agent follows, covering required actions, prohibited actions, evidence use, escalation, and the result it must return.

Agent Memory¶

Information retained across steps or sessions so an agent can use earlier facts, decisions, or results.

Agent Orchestration¶

The overall design and control of agent roles, task order, context flow, tools, and completion conditions.

Agent Output¶

The defined result an agent must return, including its content, structure, evidence, and destination.

Agent Role¶

The perspective and responsibility assigned to an agent, defining the kind of contribution it should make.

Agent Tools¶

Permitted capabilities an agent can use to retrieve information, calculate results, or act on external systems.

Agent Workflow¶

The ordered tasks, decision points, inputs, and outputs through which one or more agents complete an evaluation.

Agent-Based Simulation¶

A simulation in which multiple autonomous participants follow assigned roles and rules, producing individual and group-level behavior for analysis.

Example: A shopper agent, moderator, and brand reviewer act separately before their outputs are combined.

AI Agent¶

An AI participant that receives a defined role, goal, context, and boundaries, then performs assigned evaluation work with a specified degree of autonomy.

AI Capabilities¶

Tasks an AI system can perform adequately under stated conditions, such as generating, classifying, comparing, or summarizing content.

Example: The model can summarize interview notes and classify themes, but a researcher still verifies every quoted claim.

AI Focus Groups¶

Facilitated discussions among multiple persona agents that explore reactions to a marketing question while preserving individual viewpoints and uncertainty.

Example: Four persona agents discuss why a campaign promise feels believable to some audiences but exaggerated to others.

AI Limitations¶

Known boundaries or failure tendencies that restrict how safely and accurately an AI system’s outputs can be used.

Example: The persona cannot know actual conversion behavior, so its predicted purchase intent is labeled as synthetic.

AI Persona Testing¶

A research-support method in which language models enact defined customer profiles to evaluate marketing materials, reveal hypotheses, and guide later validation.

Example: Three defined personas independently review two landing pages, after which their hypotheses are checked in customer interviews.

Ambiguous Instructions¶

Wording that permits multiple reasonable interpretations because the requested action, scope, criteria, priority, or expected result is unclear.

Example: “Make the message stronger” leaves strength undefined and produces incompatible revisions.

Anxiety Forces¶

Uncertainty and perceived risk that make a customer hesitate before adopting a new solution.

Example: The buyer fears lost records and hidden fees during the switch.

See also: Habit Forces, Customer Circumstance

Artificial Intelligence¶

Computer systems that perform tasks associated with human reasoning, language, perception, or decision-making by applying learned patterns and explicit instructions.

Example: A research tool classifies 600 open-text comments by topic, while a marketer reviews the categories before using them.

Asset Node¶

A graph node representing a specific marketing asset and linking it to evaluations, reactions, evidence, and versions.

Audience Definition¶

A description of who will use or receive the output, guiding language, detail, and emphasis.

Example: The report is written for a marketing director who needs actions, evidence, risks, and expected impact.

Audit Trail¶

A chronological record of inputs, versions, actions, decisions, and outputs that enables later review of a research process.

Example: The project log preserves prompt v3, model settings, reviewer approval, and the final report timestamp.

Automated Comparison¶

Machine-assisted alignment and contrast of multiple assets, personas, scores, or responses using defined criteria.

Automated Reporting¶

Machine-assisted conversion of structured evaluation data into repeatable summaries, tables, visuals, and recommendations.

See also: Automated Comparison, Executive Summary

Automated Summarization¶

Machine-generated condensation of source responses into shorter findings while preserving essential meaning and traceability.

Automation Bias¶

A human tendency to favor an automated recommendation even when contrary evidence or judgment should prompt further review.

Example: A manager accepts the AI’s high trust score despite three customer recordings showing confusion.

Behavioral Consistency¶

The degree to which a persona makes compatible choices across comparable situations and repeated evaluation runs.

Example: Across website and email tests, the risk-averse persona rejects the same unsupported guarantee.

Behavioral Expert Agent¶

An expert agent that interprets choices through behavior, motivation, friction, habit, and decision-context principles.

Behavioral Segmentation¶

Grouping customers by observed actions such as usage, purchase frequency, channel choice, or response to offers.

Example: Users are grouped as weekly active, occasional, trial-only, or inactive based on recorded product use.

Brand Differentiation¶

The meaningful attributes or benefits that make a brand preferable and recognizably distinct from competitors.

See also: Competitive Comparison, Evaluation Framework

Brand Identity¶

The combined verbal, visual, and behavioral elements through which an organization presents a recognizable and coherent brand.

See also: Marketing Asset, Brand Positioning

Brand Loyalty¶

A customer’s sustained preference for a brand despite available alternatives or occasional switching incentives.

Example: A long-term subscriber stays after a small price increase because prior support has been reliable.

Brand Personality¶

The set of human-like characteristics expressed consistently by a brand, such as practical, bold, warm, or authoritative.

Brand Positioning¶

The intended place a brand occupies in a chosen audience’s mind relative to alternatives, based on relevance and meaningful difference.

Brand Promise¶

The value or experience a brand consistently commits to deliver to customers.

Brand Strategist Agent¶

An expert agent that examines whether an asset supports the intended positioning, promise, personality, and long-term brand direction.

Brand Stress Testing¶

Deliberate evaluation of a brand under difficult scenarios, skeptical audiences, and competing interpretations to expose weaknesses before launch.

Example: A skeptic persona challenges the phrase “guaranteed results” while an accessibility persona tests the signup flow.

See also: AI Focus Groups, Customer Research

Brand Tone¶

The situational variation in brand voice used to fit a particular audience, channel, topic, or emotional moment.

Brand Voice¶

The stable verbal character a brand uses across communications, expressed through vocabulary, rhythm, perspective, and style.

Budget Customer¶

A buyer who gives affordability and total cost greater weight than premium features, status, or convenience.

Example: A small-business owner chooses the basic plan after comparing total annual cost and required add-ons.

Business Impact¶

The expected effect of a finding or action on outcomes such as revenue, cost, retention, reputation, or strategic progress.

Buying Behaviors¶

Observable patterns in how customers search, compare, select, purchase, and repurchase offerings.

Example: The customer reads peer reviews, compares three plans, requests a demo, and seeks manager approval.

See also: Social Needs, Purchase Triggers

Call-to-Action Evaluation¶

Assessment of whether an action prompt is noticeable, specific, relevant, low-friction, and consistent with the customer’s next step.

Campaign Evaluation¶

Assessment of a coordinated campaign’s strategy, creative consistency, audience fit, channel execution, and cumulative effect.

See also: Video Evaluation, Social Content Evaluation

Capstone System¶

The complete course project that integrates personas, prompts, agents, evaluation, evidence, graphs, and reporting into one working process.

Channel Preferences¶

Communication and purchasing channels a customer favors for particular tasks or situations.

Example: The customer prefers email for records, chat for quick questions, and phone support for urgent problems.

Character Fidelity¶

The degree to which an agent’s responses preserve the assigned persona’s viewpoint, language, knowledge, and priorities.

Example: The persona uses its defined budget, evidence standards, and cautious tone throughout a long discussion.

Character Maintenance¶

Keeping an enacted persona’s language, priorities, knowledge, and decisions aligned with its profile throughout an interaction.

Example: The budget persona continues rejecting annual contracts after a long moderator discussion.

See also: Hallucination Reduction, Neutral Question Design

Citation Field¶

A structured-output location identifying the source or exact asset location that supports a statement.

Example: The record points to pricing-page, section 3, sentence 2 rather than naming no source.

Claim Verification¶

Checking a statement against reliable evidence to determine whether it is supported, qualified, contradicted, or unresolved.

Example: A reviewer checks “rated number one” against the cited independent ranking before approving the copy.

Clarity Score¶

A rating of how easily an intended audience can understand an asset’s main message and required action.

See also: Trust Score, Emotional Resonance

Color Evaluation¶

Assessment of a color system’s emotional associations, contrast, accessibility, cultural implications, and brand consistency.

See also: Logo Evaluation, Typography Evaluation

Comparative Benchmark¶

A reference result from another asset, version, competitor, or prior study used to interpret current performance.

Competitive Analyst Agent¶

An expert agent that compares an offer with alternatives using equivalent evidence and clearly stated comparison criteria.

See also: Brand Strategist Agent, Copywriting Expert Agent

Competitive Comparison¶

A structured assessment of an offering or asset against named alternatives using equivalent criteria and evidence.

Confidence Calibration¶

Alignment between an expressed confidence level and the actual strength, consistency, and limitations of supporting evidence.

Example: A mixed finding with limited evidence receives moderate confidence rather than a certain conclusion.

See also: Uncertainty Disclosure, Ethical Guardrails

Confidence Field¶

A structured-output location for the reviewer’s confidence level and, when required, its justification.

Example: The reviewer records moderate because two observations agree but no customer validation exists.

See also: Evidence Field, Recommendation Field

Confidence Rating¶

A reviewer’s structured estimate of how strongly the available evidence supports a finding or score.

Confirmation Bias¶

A tendency to seek, interpret, or remember information in ways that favor an existing belief or desired conclusion.

Example: The team keeps positive persona comments and dismisses negative ones because leadership favors the campaign.

Conflict Resolution¶

A documented method for addressing incompatible agent instructions, findings, or recommendations without hiding the conflict.

See also: Finding Synthesis, Expert Critique

Conflicting Instructions¶

Two or more directions that cannot all be followed together, requiring priority rules, clarification, or an explicit resolution.

Example: One line demands JSON while another demands a narrative essay, so the expected format is unresolved.

See also: Ambiguous Instructions, Instruction Hierarchy

Confusion Point¶

A specific place where wording, design, sequence, or missing information causes misunderstanding or hesitation.

Consensus Generation¶

A synthesis process that identifies supported common ground without erasing uncertainty or meaningful dissent.

Constraint Definition¶

An explicit boundary on content, behavior, evidence, scope, format, or permitted action.

Example: The reviewer may cite only supplied assets and must label any unsupported inference.

Consumer Research Agent¶

An expert agent that checks research questions, methods, evidence, and interpretations for customer-research quality.

See also: Behavioral Expert Agent, Multi-Agent System

Context Compression¶

Shortening source material while preserving the facts, distinctions, and constraints needed for the current task.

Example: A 20-page brief becomes a two-page summary that preserves prices, claims, audiences, and constraints.

Context Management¶

The deliberate selection, organization, retention, and removal of information available to an AI system during a task.

Example: The workflow retains the approved brief and removes an obsolete campaign draft before the next run.

See also: Prompt Context, Context Prioritization

Context Prioritization¶

Ranking available information so the most authoritative and task-relevant material receives attention within limited context space.

Example: The prompt labels customer transcripts as evidence and a brainstorming memo as background only.

Contradiction Detection¶

The process of locating statements or scores that cannot all be true under the same assumptions and evidence.

See also: Sentiment Pattern, Messaging Gap

Copywriting Expert Agent¶

An expert agent that critiques wording for clarity, persuasion, voice, hierarchy, and fit with the intended audience.

Credibility Gap¶

A difference between what a message claims and what the available proof or audience expectations make believable.

Credibility Score¶

A rating of how believable an asset’s claims appear given its proof, specificity, source, and tone.

See also: Purchase Intent, Relevance Score

Criterion Node¶

A graph node representing one evaluation standard and linking it to rubrics, scores, evidence, and findings.

Cross-Persona Comparison¶

A side-by-side analysis of how and why different personas respond to the same asset or question.

See also: Pattern Detection, Response Coding

Cultural Bias¶

Systematic preference for one culture’s assumptions or norms that misrepresents people from other cultural contexts.

Example: A U.S.-focused persona treats direct eye contact as universally trustworthy in an international campaign.

See also: Selection Bias, Stereotype Risk

Customer Archetypes¶

Evidence-based patterns that summarize recurring customer motivations and behaviors without claiming to represent every individual.

Example: Interview patterns yield a “careful verifier” archetype that appears across several demographic groups.

Customer Circumstance¶

The situational conditions that shape what a customer needs and which solutions are acceptable at a given time.

Example: The contract expires in 30 days while the clinic is opening a second location.

Customer Evidence¶

Recorded information from actual customers or credible observations that supports or challenges a claim about customer behavior or experience.

Example: A recording shows three customers independently stopping at the same unexplained fee.

See also: Directional Insight, AI Capabilities

Customer Interviews¶

One-to-one conversations with real customers designed to uncover experiences, motivations, language, needs, and decision processes.

Example: A researcher asks one recent buyer to reconstruct how she compared three vendors.

Customer Job¶

The progress a customer seeks in a particular circumstance, including practical, emotional, or social dimensions.

Example: A clinic manager needs to choose and deploy a compliant scheduling system before contract renewal.

Customer Journey¶

The sequence of stages and interactions through which a customer pursues a goal with an organization or category.

Example: The study traces awareness, comparison, trial, purchase, onboarding, use, and renewal.

Customer Journey Review¶

A structured assessment of journey stages and touchpoints for friction, unmet needs, emotional reactions, and improvement opportunities.

Customer Persona¶

A customer representation designed to model a customer perspective, using relevant goals, behaviors, circumstances, and evidence while avoiding unsupported stereotypes.

Example: Maya represents time-pressed clinic managers who need predictable cost, migration support, and compliance evidence.

Customer Privacy¶

Protection of customers’ control over how information about them is collected, used, shared, retained, and disclosed.

Example: Interview recordings are access-controlled, retained for a stated period, and excluded from unrelated campaigns.

Customer Progress¶

Movement from a current struggle toward a desired outcome as the customer defines it.

Example: The buyer moves from uncertain comparison to a defensible shortlist and approved choice.

Customer Research¶

Systematic collection and analysis of information about customers’ goals, experiences, behaviors, and decisions to improve business and marketing choices.

Example: Interviews show that buyers abandon checkout because delivery dates appear only after payment details.

Customer Segmentation¶

The division of a broad market into groups whose meaningful differences support distinct research, messaging, product, or service decisions.

Example: The study separates clinic managers from solo practitioners because their approval processes and support needs differ.

Customer Simulation¶

A modeled representation of customer decisions or reactions used to explore how specified conditions may influence marketing outcomes.

Example: A model explores how a cautious buyer might respond when a free trial begins requiring a credit card.

Customer Surveys¶

Standardized questions administered to a defined group to collect comparable self-reported attitudes, behaviors, or characteristics.

Example: Five hundred subscribers rate renewal intent on the same seven-point scale.

Customer Touchpoint¶

A specific interaction between a customer and the brand, product, service, employee, or channel.

Example: A renewal email becomes one interaction in the wider customer relationship.

Data Minimization¶

The practice of collecting, retaining, and exposing only the information necessary for a defined and legitimate purpose.

Example: The study records age range instead of full birth date because the exact date is not needed.

Debate Protocol¶

Rules defining roles, speaking order, evidence requirements, rebuttals, and completion conditions for an agent debate.

Decision Accountability¶

Clear assignment of responsibility to people who approve and answer for decisions informed by research or AI.

Example: The marketing director signs the launch decision rather than attributing responsibility to the AI tool.

Decision Criteria¶

Standards a customer uses to compare alternatives and judge whether an option is acceptable.

Example: The shortlist requires under-$100 pricing, single sign-on, live support, and month-to-month terms.

Demographic Segmentation¶

Grouping customers by population attributes such as age range, location, occupation, income, or household type.

Example: The survey compares results by age range and region while avoiding claims that demographics explain motivation.

Desired Outcome¶

A specific result a customer hopes to achieve, expressed independently of any particular product or solution.

Example: The team completes migration within two days with no missed appointments.

Devil's Advocate¶

A deliberately opposing role used to challenge prevailing conclusions and uncover overlooked weaknesses or assumptions.

Differentiation Score¶

A rating of how clearly an asset communicates meaningful advantages that distinguish it from alternatives.

Directional Insight¶

A preliminary indication of a likely pattern or issue that guides exploration but is not treated as conclusive customer evidence.

Example: Synthetic reviews suggest that pricing language may confuse buyers, prompting further customer testing rather than a final claim.

Disagreement Analysis¶

Examination of why reviewers differ, including differences in evidence, assumptions, criteria, or customer perspective.

See also: Minority Opinion, Moderator Neutrality

Discussion Agenda¶

An ordered set of questions and topics that keeps a group evaluation focused on its research objective.

Drift Detection¶

The process of identifying responses that no longer align with a persona’s profile or earlier established behavior.

Example: A monitor flags the budget persona after it suddenly recommends the most expensive plan for prestige.

Edge-Case Persona¶

A profile representing an uncommon but plausible situation that can expose failures hidden by typical customer cases.

Example: A buyer on slow mobile service exposes a checkout failure absent from standard desktop tests.

Email Evaluation¶

Assessment of an email’s subject line, sender trust, message hierarchy, relevance, readability, and call to action.

Emotional Drivers¶

Feelings people seek or avoid that strongly influence attention, preference, and action in a buying situation.

Example: Fear of appearing unprepared and relief from uncertainty influence the manager’s shortlist.

Emotional Job¶

The feeling a customer wants to create, preserve, or avoid while making progress in a situation.

Example: The manager wants to feel confident rather than anxious when presenting the recommendation.

Emotional Reaction¶

A reported or simulated feeling prompted by an asset, recorded separately from interpretation or behavioral intention.

Emotional Resonance¶

The degree to which a message connects with an audience’s meaningful feelings, identity, experiences, or aspirations.

Ethical AI Use¶

Application of AI in ways that respect people, prevent foreseeable harm, protect information, and communicate limitations honestly.

Example: Synthetic profiles explore early copy while real customers provide consent for later validation interviews.

Ethical Escalation¶

A defined process for raising unresolved ethical concerns to someone with authority to investigate or stop the work.

Example: A researcher pauses the study and sends a suspected discrimination issue to the responsible review lead.

See also: Decision Accountability, Customer Persona

Ethical Guardrails¶

Defined safeguards that constrain AI-supported research to reduce privacy, fairness, deception, misuse, and harm risks.

Example: The workflow blocks targeting recommendations based on disability and routes the case to human review.

Evaluation Baseline¶

The reference performance measured before a change, used to determine whether a later version improves.

Evaluation Criterion¶

A single stated standard against which an asset or response is judged.

Evaluation Framework¶

An organized approach that connects a research objective with criteria, methods, evidence, analysis, and decisions.

Evaluation Heat Map¶

A grid that uses color or intensity to reveal high and low ratings across criteria, personas, or assets.

Evaluation History¶

The chronological record of prior assets, methods, findings, scores, recommendations, and decisions.

Evaluation Instructions¶

Procedural guidance explaining which criteria reviewers apply, what evidence they record, how they assign ratings, and how they handle uncertainty.

Example: Reviewers examine clarity first, then trust, citing one observation for every rating.

Evaluation Pipeline¶

The connected stages that move an asset from intake through persona reviews, synthesis, validation, and reporting.

Evaluation Prompt¶

Instructions that define the asset, audience, question, criteria, evidence rules, and response required for an assessment.

Example: The request names the asset, audience, rubric, rating scale, evidence rule, and response fields.

Evaluation Rubric¶

A set of criteria, rating levels, and anchors used to produce consistent, evidence-based judgments.

Evaluative Research¶

Inquiry that judges how well an existing concept, experience, or marketing asset performs against defined criteria.

Example: Participants attempt checkout while researchers assess the existing page against completion and clarity criteria.

Evidence Collection¶

The deliberate capture of observations, quotations, source facts, and response details that support evaluation findings.

Evidence Field¶

A structured-output location reserved for the observation or source detail that supports a finding.

Example: The evidence value contains the exact phrase “fees may apply” and its location below the pricing table.

Evidence Node¶

A graph node containing a source observation or citation that supports or challenges a claim, score, or recommendation.

Evidence Quality¶

The strength of evidence judged by its relevance, source credibility, specificity, consistency, and independence.

See also: Supporting Evidence, Factual Critique

Evidence Triangulation¶

Comparison of multiple methods, sources, or perspectives to determine whether a finding remains credible across independent evidence.

Example: Persona outputs, customer interviews, and web analytics all point to uncertainty about cancellation.

Evidence-Based Decision¶

A choice justified by relevant evidence, transparent reasoning, uncertainty, and accountable human judgment.

See also: Implementation Effort, Improvement Hypothesis

Executive Dashboard¶

A compact visual display of key measures, findings, risks, trends, and actions for leadership review.

Executive Summary¶

A brief decision-focused account of the objective, most important findings, risks, recommendations, and limitations.

Expert Critique¶

Evaluation from a defined professional perspective that tests an asset against relevant standards and practice.

Expert Reviewer Agent¶

An agent assigned relevant professional expertise and criteria to critique an asset independently of customer personas.

Exploratory Research¶

Open-ended inquiry used to discover patterns, questions, language, or hypotheses when the problem is not yet well defined.

Example: Open interviews investigate why trial users disengage before the team decides which explanation to test.

See also: Iterative Research, Evaluative Research

Factual Critique¶

An evaluation focused on whether an asset’s statements are accurate, supportable, complete, and not misleading.

Fairness Review¶

A structured examination of whether methods or outputs disadvantage groups, rely on stereotypes, or distribute errors inequitably.

Example: Reviewers compare outputs across demographic variants and investigate an unexplained rating difference.

False Confidence¶

Unwarranted certainty in a claim, score, or recommendation despite weak evidence, high uncertainty, or known methodological limits.

Example: A report calls a finding “certain” even though it comes from one synthetic persona and no customer evidence.

Feedback Credibility¶

The degree to which feedback is believable and decision-relevant because its source, reasoning, evidence, and limitations are clear.

Example: A critique cites the exact pricing sentence, explains the concern, and labels the persona response as synthetic.

Few-Shot Examples¶

A small set of sample inputs and desired outputs included to demonstrate the intended response pattern.

Example: Two labeled input-output pairs demonstrate how to score clear and unclear cancellation language.

Finding Synthesis¶

Interpretive combination of related observations into a concise finding that preserves evidence, disagreement, and limitations.

Follow-Up Questions¶

Questions based on an earlier response that clarify meaning or explore a newly revealed issue.

Example: After a persona mentions trust, the moderator asks which exact page element changed it.

Functional Job¶

The practical task a customer is trying to complete or problem the customer needs to solve.

Example: The manager must migrate 4,000 appointments without losing records.

Functional Needs¶

Practical requirements a product or experience must satisfy for a customer to complete a task successfully.

Example: The buyer needs to export invoices, assign user permissions, and complete setup within one day.

Generative AI¶

Artificial intelligence that creates new text, images, audio, or other content by predicting plausible patterns from its training and supplied context.

Example: A model drafts five headline alternatives from a creative brief rather than selecting from a fixed list.

Goal Node¶

A graph node representing a customer goal and connecting it with personas, circumstances, barriers, and outcomes.

Graph Edge¶

A stored connection between two graph nodes, directed and labeled to express a specific relationship.

Graph Node¶

A distinct entity or record in a knowledge graph, such as one persona, asset, finding, or recommendation.

Graph Relationship¶

The meaningful association represented by an edge, such as a persona reacting to an asset or evidence supporting a finding.

Graph-Based Evaluation¶

An assessment whose personas, assets, criteria, reactions, evidence, findings, and recommendations are stored as connected graph records.

Groupthink Risk¶

The possibility that pressure for agreement suppresses doubts, alternatives, or contrary evidence in a group.

Habit Forces¶

Familiar routines, convenience, and learned behavior that keep a customer using the current solution.

Example: Staff familiarity with the old interface makes replacement feel costly despite its weaknesses.

Hallucination Detection¶

The process of identifying unsupported, invented, or contradictory statements in AI-generated content through checks against evidence.

Example: A verifier compares each factual statement with the brief and flags the invented award.

Hallucination Reduction¶

Prompt, evidence, and review practices that lower the frequency of unsupported statements in AI outputs.

Example: The prompt requires source quotations and “not provided” for missing facts, then a verifier checks each claim.

Harm Assessment¶

A structured evaluation of possible adverse effects, their severity, affected people, likelihood, and available mitigations.

Example: The team rates potential financial loss, affected customers, likelihood, and mitigation before launch.

Headline Evaluation¶

Assessment of whether a headline quickly communicates a relevant, understandable, credible, and motivating primary message.

See also: Tagline Evaluation, Call-to-Action Evaluation

Human Judgment¶

Reasoned interpretation by people who apply context, experience, ethics, and accountability to evidence and AI-generated suggestions.

Example: A researcher rejects an AI suggestion that conflicts with customer evidence and would obscure cancellation terms.

See also: Feedback Credibility, Human Oversight

Human Oversight¶

Active review and control by accountable people who can question, correct, reject, or escalate AI-supported work.

Example: The workflow pauses before publishing a report until an accountable reviewer checks claims and privacy risks.

Image Evaluation¶

Assessment of photography or illustration for relevance, authenticity, composition, representation, emotional effect, and brand fit.

Implementation Effort¶

The time, cost, skills, coordination, and technical change required to carry out a recommendation.

Improvement Hypothesis¶

A testable prediction that a specific change will improve a defined customer or business outcome for a stated reason.

Improvement Plan¶

A sequenced set of changes, responsibilities, measures, and follow-up tests for addressing evaluation findings.

Independent Review¶

An assessment completed without exposure to other reviewers’ conclusions, reducing influence and preserving distinct evidence.

Information Needs¶

Facts or explanations a customer requires to understand an offer and make a confident choice.

Example: Before buying, the manager asks for total price, migration time, security documentation, and support hours.

See also: Decision Criteria, Media Habits

A person’s voluntary agreement to participate after receiving understandable information about the research, risks, data practices, and choices.

Example: An interviewee reads how recordings will be used and freely agrees before the session begins.

Instruction Hierarchy¶

The priority order used to resolve directions from system, developer, user, and embedded content sources.

Example: A high-priority privacy rule overrides embedded page text that asks the model to reveal personal data.

International Persona¶

A profile grounded in a specific national or cross-border context, including relevant language, culture, market, and regulations.

Example: A German procurement manager evaluates translated terms, euro pricing, and local privacy expectations.

Isolated Context¶

Information kept separate for each agent to protect independent judgment and prevent cross-agent influence.

See also: Shared Context, Parallel Evaluation

Iterative Improvement¶

Repeated cycles of changing, testing, learning, and refining an asset based on evidence from each cycle.

Iterative Research¶

A repeated inquiry process in which findings from one cycle shape revisions and questions for the next cycle.

Example: After round one reveals pricing confusion, the team revises the page and tests the new version.

Jobs-to-Be-Done¶

A framework that explains why customers choose an offering by examining the progress they seek in a particular circumstance.

Example: The buyer “hires” the software to replace an expiring tool without disrupting a six-person team.

Journey Friction¶

An obstacle, delay, confusion, or effort that makes customer progress more difficult.

Example: Requiring a sales call before showing price interrupts the buyer’s comparison task.

Journey Opportunity¶

A point where a targeted change could meaningfully improve customer progress, confidence, satisfaction, or loyalty.

Example: Adding a migration checklist beside pricing could reduce uncertainty at the shortlist stage.

See also: Journey Friction, Persona Consistency

Journey Stage¶

A distinct phase in the customer’s progress, such as awareness, comparison, purchase, onboarding, use, or renewal.

Example: At comparison, the buyer needs proof and pricing; at onboarding, she needs migration guidance.

See also: Customer Journey, Customer Touchpoint

Knowledge Graph¶

A network of named entities and typed relationships that preserves how personas, assets, findings, evidence, and recommendations connect.

Landing Page Evaluation¶

Assessment of one campaign page’s message hierarchy, offer clarity, proof, friction, and primary conversion action.

See also: Website Evaluation, Product Page Evaluation

Language Model Tokens¶

Small units of text, such as words or word fragments, that a language model reads and produces when processing a prompt.

Example: The phrase “budget-friendly” may be split into several units that count against the model’s input limit.

See also: Large Language Models, Model Context Window

Lapsed Customer¶

A former buyer who stopped purchasing and may hold unresolved dissatisfaction, changed needs, or stronger alternatives.

Example: A former subscriber returns only after the company addresses the support failure that caused cancellation.

Large Language Models¶

Generative systems trained on extensive text collections to predict language sequences and perform tasks such as conversation, classification, summarization, and analysis.

Example: A marketer asks a language model to compare persona reactions and summarize the reasons for disagreement.

Leading Questions¶

Questions whose wording or assumptions encourage a particular answer and can distort research findings.

Example: “Don’t you think this guarantee is misleading?” pressures the respondent toward criticism.

Logo Evaluation¶

Assessment of a logo’s recognition, distinctiveness, legibility, relevance, versatility, and fit with intended brand meaning.

Loyal Customer¶

A repeat buyer with an established brand preference, prior experience, and greater resistance to competing offers.

Example: A five-year subscriber tolerates a small price change because prior service recovery built trust.

See also: Skeptical Customer, New Customer

Market Research¶

Systematic study of markets, customers, competitors, and demand conditions to support positioning, product, and commercial decisions.

Example: A team compares customer demand, competitor pricing, and category growth before entering a new region.

Marketing Asset¶

A brand-controlled item used to communicate or sell, such as a logo, advertisement, website, email, or product page.

Media Habits¶

Recurring patterns in when, where, and how a customer consumes information or entertainment.

Example: The buyer reads industry newsletters each morning and watches product demonstrations on a laptop.

Memorability Score¶

A rating of how likely an asset’s distinctive idea or identity is to be remembered after exposure.

Message Hierarchy¶

The deliberate order of primary, supporting, and detailed messages according to what audiences should notice and understand first.

See also: Product Messaging, Messaging Consistency

Messaging Consistency¶

Alignment of claims, terminology, emphasis, and tone across channels and stages of the customer journey.

Messaging Gap¶

Missing, unclear, or inconsistent information that prevents an audience from understanding the intended message.

Minority Opinion¶

A supported interpretation held by fewer reviewers that remains important because agreement alone does not establish truth.

Missing Data Handling¶

Instructions for identifying unavailable evidence and responding without inventing values or unsupported conclusions.

Example: When competitor pricing is absent, the response says “not provided” instead of estimating it.

See also: Citation Field, Prompt Testing

Misuse Prevention¶

Measures that reduce the likelihood that a system, persona, data set, or finding will be applied for harmful purposes.

Example: Access controls prevent a hiring team from repurposing marketing personas to screen applicants.

See also: Harm Assessment, Research Governance

Model Bias¶

Systematic distortion in AI outputs arising from training data, design choices, prompts, evaluation methods, or deployment context.

Example: The model rates identical copy differently after only the customer’s gender-coded name changes.

See also: Hallucination Detection, Automation Bias

Model Context Window¶

The limited amount of tokenized information a language model can consider during one interaction, including instructions, inputs, and prior messages.

Example: A long interview transcript is shortened because the prompt, transcript, and requested response exceed the model’s available input space.

Model Hallucination¶

Plausible-sounding content generated without adequate factual support in the supplied evidence or reliable knowledge.

Example: The reviewer invents an award that never appears in the supplied product brief.

Model Instructions¶

Directions supplied to an AI system that define its role, task, constraints, priorities, or expected form of response.

Example: The configuration says, “Use only supplied evidence, report uncertainty, and return the approved evaluation fields.”

Model Responses¶

The text or other content produced by an AI model after it processes instructions and contextual information.

Example: The saved answer contains a clarity rating, cited page text, confidence level, and recommendation.

Moderated Discussion¶

A conversation guided by a neutral facilitator who manages questions, participation, focus, and summary.

Moderator Agent¶

A neutral facilitator agent that asks questions, manages participation, and summarizes discussion without deciding the preferred answer.

See also: Persona Agent, Expert Reviewer Agent

Moderator Neutrality¶

Facilitation that avoids favoring a conclusion, participant, brand, or interpretation during discussion and synthesis.

Moderator Prompt¶

Instructions that tell a facilitator agent how to ask questions, manage participation, remain neutral, and summarize discussion.

Example: The facilitator is told to invite each persona once, probe disagreement, and summarize without taking sides.

Moment of Truth¶

A touchpoint with unusually strong influence on trust, satisfaction, or the decision to continue the relationship.

Example: The first support response after a failed migration determines whether the customer continues.

Multi-Agent System¶

A coordinated arrangement of multiple AI participants with distinct roles, contexts, or tasks that contribute to one workflow.

Needs-Based Segmentation¶

Grouping customers by the problems, required benefits, or desired outcomes they are trying to address.

Example: Buyers are grouped by migration help, compliance proof, collaboration, or lowest total cost.

Negative Persona¶

A profile of someone the offering should not target, used to prevent wasted effort and misleading broad appeal.

Example: Students seeking a free personal tool are excluded from a campaign for enterprise software.

Neutral Question Design¶

Writing questions that do not suggest a preferred answer or imply that one response is more acceptable.

Example: The moderator asks, “What stands out about the price?” rather than implying that it is confusing.

New Customer¶

A recent or prospective buyer who lacks direct experience with the brand and needs orientation, proof, and reassurance.

Example: A first-time visitor needs category explanation and proof that existing customers no longer need.

No-Code Workflow¶

A process assembled through visual interfaces and configuration rather than conventional software programming.

Open-Ended Questions¶

Questions that permit respondents to answer in their own words rather than choose from fixed options.

Example: “Tell me how you would decide between these plans” permits an answer in the customer’s own terms.

Opinion Contamination¶

A change in an independent judgment caused by seeing another participant’s view before recording one’s own.

See also: Groupthink Risk, Persona Cross-Talk

Organizational Memory¶

A durable, searchable record of findings, evidence, decisions, and lessons that remains available beyond one project or team.

Output Instructions¶

Requirements that define the response’s format, fields, length, ordering, data types, and any content that must be included or omitted.

Example: The response must return one JSON object with rating, evidence, confidence, and recommendation fields.

See also: Evaluation Instructions, Prompt Specificity

Output Schema¶

The specified fields, data types, and arrangement required in a structured response.

Example: The schema requires strings for evidence and rationale plus integers from one through five for ratings.

See also: Structured Output, Response Format

Pain Point Node¶

A graph node representing a documented customer problem and linking it to evidence, journey stages, and recommendations.

Parallel Evaluation¶

Independent assessments performed at the same stage without one reviewer seeing another reviewer’s response first.

Pattern Detection¶

The systematic identification of repeated behaviors, reactions, reasons, or relationships across evaluation responses.

Persona Acceptance Test¶

The pass-or-fail check that determines whether a profile meets stated readiness criteria.

Example: The Acceptance Test field records: “fails the profile when it recommends a product that violates its stated budget.”

Persona Agent¶

An agent instructed to evaluate an asset from one defined customer profile while preserving that profile’s viewpoint.

Persona Attitudes¶

The learned positive, negative, or mixed evaluations of relevant subjects.

Example: The Attitudes field records: “welcomes automation but distrusts unexplained recommendations.”

Persona Background¶

The relevant life, work, and market history that shapes present choices.

Example: The Background field records: “five years managing purchases for a small clinic.”

Persona Balance¶

The proportionate representation of relevant viewpoints without letting one type dominate.

Example: The Balance field records: “contains two cautious and two adventurous buyers instead of six enthusiasts.”

See also: Persona Diversity, Persona Refinement

Persona Behaviors¶

The observable actions and routines relevant to the buying situation.

Example: The Behaviors field records: “reads comparison pages and asks a peer before buying.”

Persona Beliefs¶

The ideas the represented customer accepts as true about a product, market, or situation.

Example: The Beliefs field records: “switching software will disrupt her team for several weeks.”

Persona Benchmark¶

The reference profile result used to compare later versions or model configurations.

Example: The Benchmark field records: “compares the revised profile with the approved v1.2 responses.”

See also: Persona Acceptance Test, Prompt Engineering

Persona Comparison Report¶

A report that contrasts how defined personas responded, where they agreed, and why important differences appeared.

Persona Completeness¶

The coverage of the fields needed for the intended evaluation.

Example: The Completeness field records: “includes goals, barriers, context, criteria, and evidence.”

See also: Persona Specificity, Customer Segmentation

Persona Consistency¶

The alignment among a profile’s traits, circumstances, and responses across tasks.

Example: The Consistency field records: “rejects unsupported claims in both the website and email tests.”

Persona Context¶

The situation in which the represented customer encounters the tested asset.

Example: The Context field records: “comparing plans on a phone during her commute.”

See also: Persona Narrative, Persona Demographics

Persona Coverage¶

The extent to which the profile set represents the important customer situations in scope.

Example: The Coverage field records: “adds a mobile, low-connectivity buying situation missing from the set.”

Persona Cross-Talk¶

Direct interaction among persona agents that can reveal reactions but may also blur their independent viewpoints.

Persona Demographics¶

The observable population attributes such as age range, location, or household type.

Example: The Demographics field records: “age 42, Chicago, two-person household.”

Persona Differentiation¶

The meaningful separation between this profile and other profiles in the study.

Example: The Differentiation field records: “prioritizes compliance while the second persona prioritizes setup speed.”

See also: Persona Quality, Persona Overlap

Persona Disclosure¶

A clear statement that a profile or response is synthetic, including its source, purpose, and relevant limitations.

Example: The Disclosure field records: “a concrete value relevant to the current buying decision.”

Persona Diversity¶

The meaningful variety of circumstances, needs, and viewpoints across the profile set.

Example: The Diversity field records: “includes distinct needs and circumstances rather than cosmetic demographic changes.”

Persona Documentation¶

The record of a profile’s sources, assumptions, fields, versions, and intended uses.

Example: The Documentation field records: “links every profile claim to a source or assumption label.”

Persona Drift¶

The unintended movement away from the profile’s defined traits or viewpoint.

Example: The Drift field records: “begins praising luxury features despite being defined as budget constrained.”

See also: Character Fidelity, Drift Detection

Persona Evidence Base¶

The documented research sources and assumptions supporting the profile’s characteristics.

Example: The Evidence Base field records: “six interviews and 74 coded support tickets.”

Persona Frustrations¶

The recurring sources of annoyance or dissatisfaction during a task or journey.

Example: The Frustrations field records: “support answers arrive after the purchasing deadline.”

See also: Persona Pain Points, Emotional Drivers

Persona Goals¶

The results the represented customer is actively trying to achieve.

Example: The Goals field records: “select a replacement tool before Friday.”

See also: Persona Beliefs, Persona Motivations

Persona Identity¶

The recognizable combination of traits that makes the represented customer distinct.

Example: The Identity field records: “Maya, a cautious operations manager.”

Persona Library¶

The organized collection of approved profiles available for discovery and reuse.

Example: The Library field records: “lists approved profiles by market, purpose, owner, and version.”

Persona Motivations¶

The internal or external reasons that energize a choice or action.

Example: The Motivations field records: “avoid downtime and look prepared to leadership.”

Persona Narrative¶

The brief story connecting the profile’s circumstances, needs, and behavior.

Example: The Narrative field records: “she needs a compliant tool before her current contract expires.”

Persona Needs¶

The conditions or benefits required to resolve a problem or make progress.

Example: The Needs field records: “clear migration steps and predictable total cost.”

Persona Node¶

A knowledge-graph record representing one persona and linking its profile, evaluations, reactions, and evidence.

See also: Graph-Based Evaluation, Asset Node

Persona Overlap¶

The shared characteristics that may cause two profiles to produce similar perspectives.

Example: The Overlap field records: “shares price concern with two other profiles.”

Persona Pain Points¶

The specific problems or obstacles that create difficulty in the customer experience.

Example: The Pain Points field records: “hidden fees appear only at checkout.”

Persona Preferences¶

The options or experiences the represented customer tends to favor.

Example: The Preferences field records: “prefers transparent monthly pricing and live support.”

See also: Persona Behaviors, Persona Attitudes

Persona Profile¶

The complete structured record of the customer representation.

Example: The Profile field records: “a concrete value relevant to the current buying decision.”

See also: Persona Template, Persona Identity

Persona Prompt¶

Instructions that define a persona’s profile, situation, behavioral rules, evidence limits, and response style for a simulation.

Example: The Prompt field records: “a concrete value relevant to the current buying decision.”

Persona Psychographics¶

The interests, values, attitudes, and lifestyles that help explain preferences.

Example: The Psychographics field records: “values independence and dislikes status-driven purchases.”

Persona Purpose¶

The research question and business decision the profile is intended to inform.

Example: The Purpose field records: “decide whether the page reassures first-time buyers.”

Persona Quality¶

The overall fitness of a profile for its stated research purpose.

Example: The Quality field records: “passes evidence, completeness, differentiation, and consistency checks.”

Persona Quote¶

The short first-person statement that captures characteristic language without pretending to be verbatim evidence.

Example: The Quote field records: ““Show me the full cost before I create an account.”.”

See also: Trust Disposition, Persona Scenario

Persona Realism¶

The degree to which a persona’s characteristics and responses form a plausible, evidence-grounded representation of a customer.

Example: The Realism field records: “a concrete value relevant to the current buying decision.”

See also: Simulation Fidelity, Response Variability

The evidence-based revision that makes a profile clearer, more realistic, or more useful.

Example: The Refinement field records: “replaces an assumed preference with wording found in interviews.”

Persona Reuse¶

The application of an existing profile to another compatible question after checking fit.

Example: The Reuse field records: “uses the profile again only after confirming the new product has the same decision context.”

See also: Persona Library, Persona Documentation

Persona Scenario¶

The specific situation used to place the represented customer in a realistic decision context.

Example: The Scenario field records: “her current contract expires in 30 days.”

Persona Specificity¶

The level of concrete, decision-relevant detail in a profile.

Example: The Specificity field records: “names the decision, deadline, budget, and proof required.”

Persona Template¶

The standard set of fields used to build comparable customer profiles.

Example: The Template field records: “a concrete value relevant to the current buying decision.”

Persona Test Case¶

The defined situation and expected behavioral checks used to examine profile performance.

Example: The Test Case field records: “asks the profile to identify an unsupported guarantee and cite the text.”

Persona Validation¶

The comparison of a profile and its outputs with source evidence and real-customer findings.

Example: The Validation field records: “matches its predicted concerns against five customer interviews.”

Persona Values¶

The enduring principles used to judge what matters or feels acceptable.

Example: The Values field records: “privacy and reliability matter more than novelty.”

Persona Versioning¶

The recording successive profile revisions so changes and results remain traceable.

Example: The Versioning field records: “stores v1.3 with its changed risk tolerance and test results.”

Personal Data¶

Information that identifies a person directly or can reasonably be linked with other information to identify that person.

Example: An email address and purchase history can be linked to one buyer and therefore require protection.

Persuasiveness Score¶

A rating of how effectively an asset combines relevance, evidence, emotion, and action to influence a decision.

Premium Customer¶

A buyer willing to pay more for superior quality, service, exclusivity, performance, or reduced risk.

Example: A buyer selects the higher-priced plan for priority support, audit logs, and reduced implementation risk.

See also: Budget Customer, Jobs-to-Be-Done

Price Sensitivity¶

The degree to which price changes affect a customer’s interest, choice, or willingness to buy.

Example: A $10 monthly increase moves the budget persona from the preferred plan to a competitor.

See also: Technology Comfort, Risk Tolerance

Primary Persona¶

The customer profile representing the audience whose needs most strongly guide the evaluated design or message.

Example: The redesign chiefly serves Maya, the clinic manager who owns the purchase decision.

See also: Customer Archetypes, Secondary Persona

Probabilistic Output¶

A result selected from possible continuations according to modeled likelihoods, causing reasonable responses to vary across otherwise similar runs.

Example: The same persona rates a headline 3 on one run and 4 on another, although both explanations identify unclear pricing.

See also: Model Responses, Synthetic Users

Probing Questions¶

Follow-up questions that seek greater detail, reasoning, evidence, or clarification about an earlier response.

Example: After “the price worries me,” the moderator asks, “Which part creates that concern?”

See also: Open-Ended Questions, Adversarial Questions

Product Messaging¶

The coordinated claims and explanations used to communicate a product’s audience, benefits, features, proof, and differentiation.

Product Page Evaluation¶

Assessment of whether a product page supplies the benefits, specifications, proof, price, and purchase guidance customers need.

Prompt Benchmark¶

A stable set of test cases and reference results used to compare prompt or model versions.

Example: Twenty unchanged cases provide a stable comparison for prompt v3 and prompt v4.

Prompt Chaining¶

A workflow in which one prompt’s output becomes controlled input to a later prompt.

Example: One prompt codes individual reactions; its structured output feeds a second prompt that synthesizes themes.

Prompt Clarity¶

The degree to which instructions have one understandable interpretation and use direct, consistent language.

Example: The team replaces “make it better somehow” with one direct task and an explicit success condition.

Prompt Comparison¶

A controlled test of two or more prompts on the same cases using the same evaluation criteria.

Example: Two prompt versions evaluate the same assets with the same model, settings, and rubric.

See also: Prompt Benchmark, Prompt Iteration

Prompt Context¶

Background information and source material supplied in a prompt so the model can interpret the task correctly.

Example: The prompt includes the product brief, pricing page, persona profile, and study objective before the task.

Prompt Engineering¶

The deliberate design and testing of instructions, context, examples, and output requirements to improve an AI system’s task performance.

Example: A team adds a role, evidence rule, output schema, and test cases, then keeps only revisions that improve the benchmark.

Prompt Examples¶

Sample inputs, behaviors, or outputs included to show how instructions should be applied.

Example: A sample shows an acceptable finding with a quoted sentence, rating, confidence, and bounded recommendation.

See also: Prompt Grounding, Few-Shot Examples

Prompt Failure Mode¶

A recurring way a prompt produces an incorrect, inconsistent, unsafe, or unusable response.

Example: The model repeatedly gives a rating without evidence whenever the source text is long.

Prompt Grounding¶

Requiring a response to rely on supplied evidence or identified sources rather than unsupported model-generated claims.

Example: The reviewer is required to quote the supplied page before asserting that pricing is unclear.

Prompt Iteration¶

A cycle of revising a prompt, testing it, reviewing failures, and retaining changes supported by results.

Example: After a missed citation, the team revises the evidence rule and reruns the failed case.

Prompt Objective¶

The single outcome a prompt is intended to produce, stated clearly enough to judge whether the response succeeds.

Example: The instruction begins, “Identify the two strongest barriers to signup.”

Prompt Specificity¶

The amount of concrete detail that narrows a prompt’s task, audience, evidence, constraints, and expected response.

Example: “Review this” becomes “Rate headline clarity for first-time buyers on a five-point anchored scale.”

Prompt Structure¶

The ordered arrangement of role, task, context, constraints, examples, and output directions within an instruction set.

Example: The saved prompt contains labeled Role, Task, Context, Constraints, and Output sections in that order.

Prompt Templates¶

Standard prompt structures containing fixed instructions and clearly marked fields for variable content.

Example: A locked instruction block surrounds editable fields for audience, asset, question, and rubric.

See also: Reusable Prompts, Prompt Chaining

Prompt Test Case¶

A documented input, context, expected behavior, and pass condition used to evaluate a prompt.

Example: One case supplies an unsupported guarantee and passes only if the model flags and quotes it.

Prompt Testing¶

Running defined cases against a prompt to find failures and measure whether revisions improve performance.

Example: The team runs ten fixed cases after every revision and records pass or fail for each requirement.

Prompt Variables¶

Named placeholders whose values change between runs while the surrounding prompt remains stable.

Example: The template substitutes {{persona}}, {{asset}}, and {{rubric}} for each evaluation run.

Prompt Versioning¶

Recording prompt revisions and their results so teams can reproduce, compare, and restore tested instructions.

Example: The log links prompt v2.4 to its changed evidence rule and benchmark results.

Psychographic Segmentation¶

Grouping customers by values, attitudes, interests, identity, or lifestyle rather than observable demographics alone.

Example: One group prioritizes independence and novelty; another values predictability and expert guidance.

See also: Demographic Segmentation, Needs-Based Segmentation

Pull Forces¶

Attractive benefits and anticipated progress that draw a customer toward a new solution.

Example: A guided migration service and transparent pricing attract the buyer to the new vendor.

Purchase Barriers¶

Concerns or obstacles that delay, prevent, or redirect a customer’s buying decision.

Example: An annual commitment, unclear cancellation terms, and missing security evidence delay signup.

Purchase Intent¶

A customer’s stated likelihood or willingness to buy, treated as an attitude rather than proof of future behavior.

Purchase Triggers¶

Events or conditions that move a customer from passive interest toward an active buying decision.

Example: An expiring contract and a limited migration window prompt the team to begin vendor selection.

Push Forces¶

Dissatisfaction or changed circumstances that make a customer’s current solution less acceptable.

Example: Repeated outages make the current scheduling tool unacceptable.

Qualitative Research¶

Research that examines meanings, motivations, language, and experiences through nonnumeric evidence such as interviews, observations, and open responses.

Example: Eight interviews reveal that customers describe the onboarding process as “uncertain” rather than merely “slow.”

Quantitative Research¶

Research that measures variables numerically and analyzes patterns using counts, percentages, comparisons, or statistical methods.

Example: A survey finds that 38 percent of 500 respondents recognize the new tagline.

See also: Qualitative Research, Traditional Focus Groups

Rating Scale¶

An ordered set of response choices used to represent different levels of a measured quality.

Rating Scale Output¶

A structured response that records a selected rating together with its scale and supporting rationale.

Example: The record stores clarity: 2, the anchor “frequent confusion,” and supporting page text.

Rationale Field¶

A structured-output location explaining why a rating, finding, or recommendation follows from the evidence.

Example: The explanation connects the low trust rating to an unexplained renewal condition in the asset.

Reaction Node¶

A graph node representing a persona’s recorded emotional, cognitive, or behavioral response to an asset.

See also: Pain Point Node, Criterion Node

Real Customer Validation¶

Testing a synthetic finding with appropriate evidence from actual customers before treating it as decision-ready.

Example: Five recruited buyers confirm that the pricing label identified by synthetic personas is confusing.

See also: Synthetic Data, Evidence Triangulation

Recommendation¶

A proposed action tied to a documented finding, intended outcome, and supporting evidence.

See also: Risk Likelihood, Recommendation Priority

Recommendation Field¶

A structured-output location for an action proposed in response to a documented finding.

Example: The response proposes moving the total monthly price beside the primary signup button.

Recommendation List¶

An ordered collection of proposed actions, each connected to evidence, expected impact, priority, and ownership.

See also: Evaluation Heat Map, Improvement Plan

Recommendation Node¶

A graph node representing a proposed action and linking it to supporting findings, priority, effort, and expected impact.

Recommendation Priority¶

The relative order in which proposed actions should be considered based on impact, urgency, evidence, risk, and effort.

Reflection Prompt¶

An instruction asking an agent to reconsider its reasoning, evidence, assumptions, or consistency before finalizing.

Example: Before finalizing, the reviewer checks whether every claim has evidence and whether the persona remained consistent.

Relationship Type¶

A controlled label that defines what a graph edge means, such as SUPPORTS, EVALUATES, REACTS_TO, or ADDRESSES.

See also: Evidence Node, Organizational Memory

Relevance Score¶

A rating of how closely an asset addresses the intended audience’s situation, priorities, and desired outcomes.

Repeatable Evaluation¶

An assessment whose documented inputs, prompts, criteria, and procedures can be applied again in a consistent way.

Example: The saved asset, prompt version, model setting, rubric, and run procedure allow the test to be run again.

Reproducible Results¶

Findings that can be obtained again by another evaluator using the documented materials, settings, and procedure.

Example: A second analyst follows the documented procedure and obtains the same leading themes and similar scores.

Research Assumptions¶

Conditions treated as true for planning or interpretation even though they may not have been directly established by evidence.

Example: The team records its unverified belief that visitors already understand the product category.

Research Constraints¶

Limits on a study imposed by available time, budget, data, tools, access, ethics, or method.

Example: The study has one week, no customer email list, and a maximum recruitment budget of $3,000.

Research Cost¶

The money, labor, tools, incentives, and opportunity costs required to plan, conduct, analyze, and report a study.

Example: The budget includes recruitment, incentives, moderator time, transcription, software, and analysis.

Research Governance¶

Roles, standards, approvals, documentation, and controls that make research quality, ethics, and accountability manageable.

Example: A policy names the study owner, approval steps, evidence standard, retention period, and escalation contact.

Research Objectives¶

Clear statements of what a study intends to learn, evaluate, compare, or decide.

Example: The brief states that the study will identify the two strongest barriers to trial signup.

Research Questions¶

Specific questions a study is designed to answer through evidence rather than assumption or preference.

Example: “Which proof points make first-time buyers trust the guarantee?” directs the study toward a specific answer.

Research Reliability¶

The degree to which a method produces sufficiently consistent measurements or findings when repeated under comparable conditions.

Example: Two trained reviewers applying the same coding guide assign nearly all comments to the same themes.

See also: Research Validity, Repeatable Evaluation

Research Sample Size¶

The number of participants, observations, or simulated profiles included in a study and considered when interpreting its evidence.

Example: The report states that 12 interviews support qualitative themes but not population estimates.

See also: Customer Surveys, Research Turnaround Time

Research Scope¶

The defined boundaries of a study, including included audiences, materials, questions, contexts, and time period.

Example: The study covers U.S. first-time buyers, the mobile landing page, and the signup decision, but not retention.

See also: Research Objectives, Research Assumptions

Research Transparency¶

Clear disclosure of a study’s purpose, methods, sources, assumptions, limitations, and use of AI.

Example: The report lists the model, prompt version, synthetic sample, assumptions, and validation status.

See also: Ethical AI Use, Informed Consent

Research Turnaround Time¶

The elapsed time from defining a study through collecting, analyzing, and communicating its findings.

Example: The team delivers a concept-test report four business days after receiving the final assets.

Research Validity¶

The degree to which evidence and methods support the interpretation or decision a study claims to support.

Example: A checkout test measures whether customers can complete checkout, not whether they remember the brand a week later.

Response Aggregation¶

Mechanical or analytical combination of multiple responses into counts, distributions, summaries, or composite results.

Response Coding¶

The assignment of consistent labels to portions of qualitative responses so similar evidence can be grouped and analyzed.

Response Consistency¶

The degree to which repeated outputs follow the same role, criteria, format, and reasoning expectations under comparable inputs.

Example: Across five runs, the persona identifies the same barrier and uses the required fields despite wording changes.

Response Format¶

The presentation form required for an answer, such as a table, list, narrative, or structured record.

Example: The executive result is requested as a two-column table followed by three prioritized actions.

Response Variability¶

Differences among outputs produced from the same or similar inputs because language model generation is probabilistic.

Example: Across five runs, a persona uses different wording but repeatedly flags the same hidden renewal condition.

Responsible AI¶

The design and use of AI with attention to accuracy, fairness, privacy, transparency, safety, and human accountability.

Example: The team minimizes personal data, tests for unfair patterns, discloses simulation, and requires human approval.

Reusable Prompts¶

Tested instructions written with stable language and replaceable inputs for repeated use across compatible evaluations.

Example: The same tested instruction evaluates five landing pages by changing only approved variables.

Reviewer Independence¶

Protection of each reviewer’s initial judgment from pressure, imitation, and prior exposure to others’ conclusions.

Reviewer Prompt¶

Instructions that tell an evaluator what asset to assess, which criteria to apply, and what evidence to report.

Example: The reviewer must score the landing page, cite exact wording, and propose one evidence-linked revision.

See also: Moderator Prompt, Skeptic Prompt

Risk Assessment¶

A structured examination of what could go wrong, whom it could affect, how likely it is, and how serious it would be.

Risk Likelihood¶

The estimated chance that an identified risk will occur under stated conditions.

Risk Severity¶

The magnitude of harm or negative business consequence if an identified risk occurs.

Risk Tolerance¶

The amount of uncertainty or possible loss a customer accepts when considering an unfamiliar choice.

Example: The cautious buyer refuses a new vendor without references and a reversible monthly contract.

Role Definition¶

A clear statement of the perspective, expertise, authority, and responsibilities assigned within a prompt.

Example: The instruction says, “You are a neutral moderator, not a brand advocate or final decision-maker.”

Round-Table Discussion¶

A structured exchange in which several agents compare perspectives after completing their initial independent reviews.

See also: Independent Review, Moderated Discussion

Rubric Completeness¶

The degree to which a rubric covers every criterion needed to answer the stated evaluation objective.

Rubric Dimension¶

A distinct aspect of performance that an evaluation rubric measures, such as clarity, trust, or relevance.

See also: Evaluation Rubric, Rubric Completeness

Rubric Fairness¶

The degree to which rubric criteria and scoring avoid irrelevant disadvantage, stereotypes, and unequal standards.

Scalable Research Process¶

A documented method that can handle more assets, personas, or studies without proportional growth in cost, time, or inconsistency.

Scale Anchors¶

Plain-language descriptions of what specific points on a rating scale mean and what evidence supports them.

Score Normalization¶

The conversion of scores from different scales or distributions into a common form for valid comparison.

See also: Weighted Criteria, Confidence Rating

Scoring Guidance¶

Instructions that explain how reviewers should choose ratings, handle uncertainty, and cite supporting observations.

Scoring System¶

A defined method for converting evidence-based judgments into comparable numbers, categories, or ordered ratings.

Secondary Persona¶

A relevant customer profile considered after the primary audience because its needs are important but less central.

Example: The accountant’s review needs remain important but do not override the clinic manager’s core workflow.

Selection Bias¶

Systematic distortion caused when the included people, personas, examples, or evidence do not adequately represent the intended population.

Example: Testing only loyal subscribers makes the new-customer onboarding page appear easier than it is.

Sensitive Attributes¶

Personal characteristics such as health, race, religion, disability, or sexuality that can create heightened privacy or discrimination risks.

Example: A health condition is excluded from the persona because it is unnecessary for testing the pricing message.

See also: Personal Data, Data Minimization

Sentiment Pattern¶

A recurring distribution or change in positive, negative, or mixed reactions across audiences, assets, or journey stages.

Sequential Evaluation¶

Assessments performed in an ordered series where a later step may use an earlier step’s result.

Shared Context¶

Information deliberately made available to multiple agents so they work from the same relevant facts and materials.

Simulation Fidelity¶

The degree to which a simulation preserves the relevant characteristics and behavior of the real situation it represents.

Example: A checkout simulation includes the real price, device layout, and cancellation terms because those details affect the decision.

Skeptic Agent¶

An agent tasked with challenging claims, evidence, and assumptions to reveal weaknesses that agreeable reviewers may miss.

Skeptic Prompt¶

Instructions that direct an agent to challenge claims, inspect risks, and identify unsupported assumptions without becoming indiscriminately negative.

Example: The skeptic is instructed to challenge guarantees, missing proof, hidden conditions, and optimistic assumptions.

Skeptical Customer¶

A customer who requires strong evidence and actively questions claims, motives, risks, and hidden conditions before accepting an offer.

Example: The buyer refuses “industry leading” until the page supplies an independent source and comparison method.

Assessment of social posts for platform fit, clarity, authenticity, engagement value, brand voice, and response risk.

The way a customer wants to be perceived by others or relate to a relevant group.

Example: The manager wants leadership to see the purchase as careful and competent.

Desired effects on belonging, recognition, status, or relationships that influence a customer’s choice.

Example: The manager wants colleagues to view her recommendation as careful and professionally credible.

Stereotype Risk¶

The possibility that a persona or output reduces a group to oversimplified traits and produces unfair or misleading conclusions.

Example: A profile assumes every retired customer dislikes technology without supporting research.

Structured Output¶

A response organized into named fields or a fixed schema so results can be checked and compared.

Example: Every review returns the same four named fields instead of an unstructured paragraph.

Supporting Evidence¶

Specific recorded information that directly justifies a finding, rating, risk, or recommendation.

Switching Forces¶

The combined pressures that push customers from a current solution, pull them toward another, or hold them back.

Example: Poor support pushes the buyer away, easier migration pulls her forward, and contract anxiety holds her back.

See also: Customer Progress, Push Forces

Synthesis Prompt¶

An instruction asking an agent to combine multiple findings into a coherent account while preserving important differences.

Example: The synthesizer combines shared pricing concerns while preserving one persona’s contrary view and its evidence.

Synthetic Data¶

Artificially generated records designed to resemble useful properties of real data without necessarily describing actual individuals.

Example: Artificial purchase records preserve broad spending patterns without reproducing any customer’s actual history.

Synthetic Personas¶

Structured, AI-operable customer profiles that combine relevant traits, goals, circumstances, and behaviors for repeatable simulated evaluation.

Example: A profile combines a price-sensitive owner’s goals, constraints, purchase criteria, and media habits for repeated tests.

Synthetic User Research¶

Structured inquiry using simulated customer representations to produce early directional findings that remain subject to validation with real people.

Example: Simulated reactions reveal a possible trust problem that the team adds to its next real-customer interview guide.

See also: AI Persona Testing, Customer Simulation

Synthetic Users¶

AI-generated representations of people used to explore possible customer reactions without claiming that the representations are real research participants.

Example: Twenty simulated shoppers review an early concept before any claim is tested with recruited customers.

System Prompt¶

High-priority instructions that establish an AI system’s role, operating rules, and behavior across an interaction.

Example: A high-priority instruction states, “Act as a neutral research moderator and never invent customer evidence.”

Tagline Evaluation¶

Assessment of a tagline’s clarity, memorability, distinctiveness, credibility, adaptability, and connection to the brand.

Task Definition¶

A clear statement of the work to perform, including the object of analysis and desired result.

Example: The task reads, “Find wording that could make cancellation terms difficult to understand.”

See also: Role Definition, Audience Definition

Technology Comfort¶

A customer’s readiness and confidence when learning or using digital tools and automated experiences.

Example: A confident user explores advanced settings alone, while a hesitant user requests guided setup.

Theme Identification¶

The process of developing broader recurring ideas from related coded observations while retaining links to source evidence.

Traditional Focus Groups¶

Moderated conversations with recruited human participants who discuss a topic together and provide qualitative evidence about their perceptions.

Example: A human moderator leads eight recruited customers through a ninety-minute discussion of package designs.

Trust Disposition¶

A person’s general tendency to trust or question organizations, claims, technology, and other people.

Example: One persona accepts verified customer reviews readily, while another demands independent documentation.

Trust Score¶

A rating of how safe, honest, reliable, and credible an asset appears to the intended audience.

Trust Signal¶

A visible cue, such as transparent terms, credible proof, or recognizable endorsement, that reduces perceived uncertainty.

Turn Taking¶

The controlled order in which participants contribute, preventing interruption and preserving an interpretable discussion record.

See also: Debate Protocol, Discussion Agenda

Typography Evaluation¶

Assessment of type choices for legibility, hierarchy, personality, accessibility, and consistency across formats.

Uncertainty Disclosure¶

Explicit communication of what is unknown, variable, assumed, or weakly supported in an analysis or recommendation.

Example: The summary states that simulated reactions suggest a risk but cannot estimate its population frequency.

Use Context¶

The immediate environment, device, time pressure, and activity surrounding a customer’s interaction with an asset.

Example: The manager compares prices on a phone during a commute and later shares screenshots with the owner.

User Prompt¶

The immediate request, question, or source material submitted for the AI system to address within higher-priority rules and available context.

Example: The marketer asks, “Compare these two headlines against the approved clarity rubric.”

Value Proposition¶

A concise explanation of the customer benefit an offering provides, for whom, and why it is preferable to alternatives.

Value-Based Segmentation¶

Grouping customers by their economic value to the organization or by the value they seek from an offer.

Example: The team distinguishes high-lifetime-value accounts from low-frequency buyers for service planning.

Video Evaluation¶

Assessment of a video’s opening, narrative, pacing, sound, visuals, comprehension, brand fit, and call to action.

Visual Design Critic¶

An expert agent that examines hierarchy, composition, color, typography, imagery, and usability in a visual asset.

Visual Identity¶

The recognizable system of logos, colors, typography, imagery, and layout that visually represents a brand.

Website Evaluation¶

Assessment of a website’s messaging, navigation, usability, trust, accessibility, and ability to support visitor goals.

Weighted Criteria¶

Evaluation criteria assigned different influence on the total result according to their importance to the decision.

Workflow Automation¶

The use of configured rules and tools to execute repeatable process steps with limited manual handling.

See also: No-Code Workflow, Evaluation Pipeline