Brand Architect

When to use this

Use when a brand needs to be captured as a working model rather than described — so that a person or an AI system can make new decisions for it without going back to the people who built it. Typical cases: documenting a brand that grew without documentation, reconstructing one from its artifacts, onboarding an agency or a new team, or giving an AI system enough of the brand to generate work that stays recognizable.

Use it when the brand exists and needs to be understood. The rest of the system builds a brand; this one models one.

Use Brand Review instead if the question is whether the brand is any good. Brand Architect describes what is true, including where the brand is weak; it does not issue a verdict. Use Brand Governance instead if what you need is decision rights and approval routes rather than a model of the brand itself — though BRAND.md is a reasonable place for governance to point.

Inputs

Whatever the brand actually consists of: guidelines, live site, product interfaces, campaigns, social posts, decks, sales material, packaging, photography, typography, color and design tokens, customer communication, founder statements, research, and competitive material.

Ask for the material rather than working from what you can reach. A model built only on the public website will describe the marketing, not the brand, and it will be confidently wrong about how the brand behaves in product, support, and failure — which is where character actually shows.

Ask which parts of the brand the team considers settled and which they consider unresolved. Both belong in the model, labeled differently.

If a section of the model has no supporting material, leave it unknown. Do not infer a brand's purpose, values, or audience from its visual style.

Mission

Create a BRAND.md that captures the smallest complete model of a brand required to understand it, make decisions for it, reproduce it, critique work against it, and extend it into situations that were not explicitly documented.

BRAND.md is not a scrapbook, moodboard, brand-book digest, or collection of adjectives. It is the brand's operational memory.

A successful file should allow a capable person or AI system to answer:

• What is this brand?
• Why does it exist?
• Who is it for?
• What does it promise?
• Why should anyone believe it?
• What makes it meaningfully different?
• What should it feel like?
• How should it sound?
• How should it look?
• How should it behave?
• What must remain recognizable?
• What can change?
• What should never happen?
• How should a new decision be made when the documentation does not contain an exact example?
• Which parts are facts, which are interpretations, and which remain unknown?

The quality bar is not completeness by volume. It is useful truth compressed into a system that can generate correct future decisions.

Operating model

Use this sequence:

Inventory → Evidence → Context → Pattern extraction → Interpretation → Brand model → Rules → Examples → Stress test → BRAND.md → Reproduction audit

Do not jump directly from source material to prose.

Inventory

Identify everything available before interpreting it. Possible inputs include brand guidelines, websites, product interfaces, campaigns, social posts, decks, sales material, product copy, packaging, photography, illustration, typography, logos, color systems, motion, customer communication, founder statements, research, competitive material, source code, and design tokens.

Record what was actually inspected. Do not imply that the entire brand was understood if only a subset was available.

Evidence

Separate:

• Confirmed: directly supported by source material or observed behavior.
• Strong inference: supported by repeated evidence but not explicitly stated.
• Working interpretation: useful model that remains revisable.
• Unknown: the material does not establish the answer.
• Conflict: two credible signals disagree.

Never turn inference into fact through confident writing.

Context

Understand the environment in which the brand operates: business model, category, market, audience, buyer/user distinction, competitors, alternatives, cultural context, product maturity, distribution, business constraints, regulatory or accessibility constraints where relevant, existing reputation, and strategic ambition.

Do not invent missing context. If context materially affects a conclusion and is unavailable, preserve the uncertainty.

Pattern extraction

Look for repeated behavior rather than isolated examples.

For language inspect sentence length, rhythm, vocabulary, verbs, nouns, specificity, abstraction, humor, confidence, emotional range, claims, calls to action, punctuation, point of view, technicality, metaphor, and forbidden patterns.

For visual identity inspect composition, hierarchy, scale, density, spacing, typography, color relationships, image treatment, shape language, iconography, illustration, motion, contrast, repetition, asymmetry, and distinctive assets.

For behavior inspect what the brand prioritizes, refuses, how it responds, how it handles uncertainty, how it treats users, how it explains difficult things, how it behaves under pressure, and how it behaves when something goes wrong.

Repeated behavior is usually more reliable than a single declarative statement.

Interpretation

Turn patterns into meaning.

Do not stop at “uses large type.” Ask what the scale accomplishes, what it suggests about the brand's character, where that behavior should remain true, and what would be a false application of the same rule.

Brand model

Construct the underlying system before writing the final file. At minimum establish essence, purpose, strategic context, audience, positioning, value proposition, differentiation, proof, personality, emotional territory, voice, tone, messaging hierarchy, visual language, distinctive assets, design behavior, brand world, behavioral principles, application behavior, boundaries, decision rules, and evidence/confidence.

Rules

Convert important observations into decisions.

A useful rule explains:

Behavior → Meaning → Reason → Context → Boundary

Avoid rules that merely say what to do without explaining why.

Examples

Examples should demonstrate the rule rather than replace it. Use strong examples, weak examples, contextual examples, edge cases, and before/after comparisons where available.

Never fabricate an example and present it as an existing brand artifact. If an example is generated to demonstrate a rule, label it as generated.

Stress test

Before finalizing, attempt to use the emerging model to make new work: a short brand statement, a homepage or campaign direction, a product/interface moment, a social/editorial message, a response to a difficult situation, and a decision where two brand principles compete.

If the outputs feel inconsistent, the model is incomplete.

Final BRAND.md

Write the canonical file only after the model survives the stress test. Prioritize durable principles over temporary implementation details.

Reproduction audit

Read BRAND.md as if you had never seen the original brand. Ask whether it gives enough information to make a new, recognizably on-brand decision without access to the original team. If not, improve the document.

Outcome orientation

The purpose of BRAND.md is not documentation for documentation's sake. It should improve consistency, decision speed, recognition, creative quality, strategic coherence, handoff quality, production quality, future brand extensions, AI-assisted creation, critique, and governance.

For important rules, explain the intended effect. Do not claim that a rule increases conversion, trust, recall, or revenue unless evidence supports that claim.

Evidence and uncertainty

Maintain epistemic discipline throughout the work.

Use the same five classes the rest of the system uses, so evidence keeps its meaning when it moves between skills.

Class	Meaning
Observed	Directly seen in the source material or in brand behavior
Reported	Stated by a brand representative or source
Inferred	Reasoned from repeated signals, not itself observed
Assumed	A necessary working assumption, unverified
Unknown	Material question the evidence does not establish

Conflict is not a sixth class. It is a flag on a pair of claims: two credible signals disagree, or stated strategy contradicts observed behavior. Record both claims with their own classes and mark the conflict — resolving it by picking the more convenient one is how a brand model acquires false confidence.

When useful, include evidence notes beside important strategic conclusions. Do not overload the final file with research bureaucracy; preserve enough provenance to prevent false certainty.

Judgment framework

Evaluate the resulting model through seven lenses:

Truth

Does the model describe what the brand actually is?

Meaning

Does it explain why the brand behaves this way?

Distinction

Does it identify what separates this brand from category convention?

Recognition

Does it preserve the cues through which the brand can be identified?

Coherence

Do strategy, language, visual identity, behavior, and applications reinforce one another?

Generativity

Can the model produce new work rather than only explain existing work?

Resilience

Can it guide decisions when the exact situation was never documented?

A beautiful document that fails generativity or resilience is incomplete.

Brand model construction

Brand essence

Define the central idea without collapsing it into an empty slogan. Capture what the brand fundamentally offers, the role it plays in people's lives, the change it seeks to create, the emotional or cultural territory it occupies, and what makes that territory credible.

Avoid generic words such as innovative, premium, simple, human, bold, or trusted unless they are made specific through evidence and behavior.

Purpose

Separate business purpose, customer value, and broader ambition. Do not force a grand social mission where the evidence does not support one.

Positioning

Document category, audience, problem/job, promise, differentiation, alternatives, proof, reasons to believe, and non-choices. A positioning statement is useful only if it creates a meaningful choice.

Personality

Do not use personality as an adjective dump. Define characteristic behavior, emotional range, social posture, relationship to authority, relationship to expertise, relationship to humor, relationship to urgency, and relationship to imperfection.

For each major trait, explain what it means in behavior.

Tensions

Strong brands often contain productive tension. Examples include serious + playful, technical + accessible, restrained + expressive, intimate + ambitious, familiar + strange. Document how the tension is resolved. Do not flatten the brand into one mood.

Verbal identity

Document the voice as a behavioral system: point of view, vocabulary, sentence rhythm, specificity, confidence, emotional range, technical language, humor, metaphor, punctuation, calls to action, claims, and forbidden patterns.

Voice vs tone

Voice is the durable character. Tone is the contextual expression.

Document how the voice changes across marketing, product, support, error states, announcements, serious situations, celebration, apology, and technical explanation. The brand should remain recognizably itself while adapting to context.

Messaging hierarchy

Distinguish:

1. master idea
2. core promise
3. supporting messages
4. proof
5. product-specific messages
6. calls to action

Do not allow every page to invent a new positioning.

Visual identity

Document the visual system as behavior rather than inventory.

Logo

Capture role, recognition value, required integrity, acceptable flexibility, and prohibited manipulation.

Typography

Do not only list fonts. Explain hierarchy, scale relationships, density, emphasis, rhythm, display vs utility behavior, and how type contributes to personality.

Color

Document roles, relationships, hierarchy, contrast, semantic usage, emotional function, distinctive use, and boundaries.

Composition

Explain alignment, rhythm, whitespace, density, scale, cropping, asymmetry/symmetry, and focal hierarchy.

Imagery

Define subject matter, framing, lighting, texture, treatment, human presence, realism vs abstraction, and emotional posture.

Motion

Where relevant, document pace, easing character, transition behavior, emphasis, restraint, and when motion should not be used.

Distinctive assets

Identify elements that carry recognition beyond the logo. For each one explain what it is, why it matters, where it works, how flexible it is, and what would weaken it.

Design behavior

This section matters especially when the brand appears in products. Document how the brand behaves in navigation, hierarchy, controls, empty states, loading, errors, success, onboarding, dashboards, dense information, responsive layouts, accessibility, motion, and interaction feedback.

Do not turn the brand document into a component library. Describe the behavioral principles that a design system should implement.

Brand world

Capture the broader world the brand inhabits: cultural references, symbols, recurring objects, rituals, narrative patterns, spatial qualities, material qualities, sonic associations, and temporal qualities where relevant.

Only include patterns supported by evidence or explicitly label them as interpretations.

Brand behavior

A brand is also how it acts. Document principles for making promises, explaining uncertainty, responding to mistakes, treating customers, handling criticism, speaking about competitors, announcing change, handling sensitive situations, and balancing confidence with honesty.

Behavior should reinforce the same character expressed through words and visuals.

Recognition system

Identify the cues that allow recognition when obvious identifiers are removed.

Test logo removed, name removed, color removed, typography changed, image changed, and context removed.

Ask: What remains?

Rank distinctive cues by confidence and importance. Do not claim distinctiveness merely because something is unusual. Distinctiveness requires evidence that the cue meaningfully separates the brand from alternatives.

Application rules

Explain how the system changes across contexts without losing its identity. At minimum consider, where relevant, website, product, social, campaign, sales, presentation, editorial, physical, internal communication, and partner communication.

For each major application ask: What remains constant? What adapts?

Boundaries

Separate:

Never

Behavior that materially contradicts the brand.

Avoid

Behavior that is usually harmful but may have exceptional contextual justification.

Flexible

Areas where local judgment is expected.

Experimental

Areas where deliberate exploration is encouraged.

This prevents the brand from becoming either chaotic or rigid.

Brand decision rules

When a new decision is not explicitly documented, the user should be able to reason from principles.

A good decision rule answers:

1. What are we trying to preserve?
2. What are we trying to achieve?
3. Which brand principle applies?
4. What evidence matters?
5. What tradeoff exists?
6. What would make the decision feel off-brand?
7. How can the decision be tested?

When principles conflict, prefer the choice that best preserves strategic truth, recognition, audience relevance, clarity, and craft, unless the brand's actual strategy provides evidence for another priority.

Failure modes

Actively prevent adjective soup, brand-book compression, false certainty, generic differentiation, visual inventory without logic, voice without behavior, rules without rationale, examples without principles, over-documentation, under-documentation, static brand thinking, AI voice contamination, invented history, and decorative strategy.

Working method

Work in explicit passes:

1. inventory
2. evidence map
3. pattern map
4. contradiction map
5. model
6. rules
7. examples
8. stress test
9. correction
10. compression
11. final audit

Compression rule

Remove anything that does not improve understanding, decision-making, reproduction, or governance.

Decision record

For consequential interpretation decisions, preserve:

Decision	Evidence	Interpretation	Alternatives	Risk	Confidence

Use this internally while constructing the document. Do not expose unnecessary process in the final BRAND.md; retain only information that improves future judgment.

Output

Produce exactly one canonical artifact, BRAND.md, following the section architecture in references/brand-md-spec.md. Follow that structure rather than inventing a parallel one, so that two BRAND.md files from different projects stay comparable.

The document should be structured, readable, searchable, and useful to both humans and AI systems. Use stable headings. Prefer short paragraphs, tables, bullets, and examples where they improve retrieval. Avoid walls of prose.

Do not force irrelevant sections into a brand. If a section genuinely does not apply, state why it is not applicable rather than inventing content.

Verification

Before delivering BRAND.md, perform all of the following.

Truth test

Can every consequential factual claim be traced to supplied evidence?

Specificity test

Could this description apply equally well to five unrelated brands? If yes, make it more specific or remove it.

Contradiction test

Does any section contradict another?

Recognition test

What survives if logo, name, and category context are removed?

Generativity test

Can the document produce credible new brand decisions?

Edge-case test

Can it guide a decision that is not directly documented?

Tone test

Does the document itself remain clear and disciplined rather than sounding like generic branding copy?

Compression test

Can anything be removed without reducing the model's usefulness?

Reproduction test

Give the BRAND.md to a fresh agent with no original brand material and ask it to produce a short brand message, a visual direction, a product interface direction, a campaign idea, and a difficult customer response. Compare the outputs against the source brand. If the outputs drift, identify which missing rule caused the drift and strengthen the document.

Fresh-reader test

Someone unfamiliar with the project should be able to understand the brand's core logic without a verbal explanation from the original author.

Completion criteria

BRAND.md is complete only when:

• the brand can be explained without generic adjective soup
• strategic context is clear
• positioning is meaningful and differentiated
• audience and relevance are understood
• personality is expressed through behavior
• voice can generate new writing
• visual identity can generate new design decisions
• design behavior is understandable beyond individual components
• distinctive assets are identified
• recognition does not depend exclusively on the logo
• brand behavior is explicit
• boundaries are clear
• decision rules work outside documented examples
• evidence and uncertainty are distinguished
• contradictions are surfaced rather than hidden
• the document can be used by humans and AI
• the document has survived reproduction testing
• no material claim is invented
• no placeholder remains
• no unnecessary documentation obscures the important decisions

The final standard is:

BRAND.md should not merely describe the brand. It should make the brand easier to understand, easier to protect, easier to extend, and harder to accidentally dilute.

Handoff

BRAND.md is the model, not a substitute for the decisions that produced it. Where it records a position, a name, a verbal system, or an identity, the reasoning behind those lives in the artifacts of Brand Strategy, Brand Naming, Verbal Identity, and Identity System — point to them rather than restating their conclusions as if they had no history.

Brand Critic and Brand Review use BRAND.md as the standard work is judged against; anything left unknown here is something they cannot hold the work to, so the unknowns matter as much as the rules. Brand Governance takes the boundaries and decision rules and assigns owners to them. Where the model surfaces a contradiction between stated strategy and observed behavior, that is a Brand Review finding, not something to smooth over in the document.