Skip to content

DOMAIN:CREATIVE — BRAND VOICE & MESSAGING FRAMEWORKS

OWNER: rick ALSO_USED_BY: jouke (Content), dinand (Content), benjamin (Content), dima (Intake — public-facing), faye (PM Alpha), sytske (PM Beta) UPDATED: 2026-03-28 PREREQUISITE: copywriting-craft.md (writing fundamentals), ux-writing.md (applied voice in UI)


VOICE VS. TONE

RULE: voice and tone are different things. Confusing them is the most common brand copy mistake.

Voice is the brand's consistent personality. It does not change between channels, campaigns, or contexts. It is WHO the brand is.

Tone is how the voice adapts to context. A doctor speaks differently at a funeral than at a birthday party, but they are still the same person. Tone shifts. Voice does not.

EXAMPLE: - Voice attribute: "Confident" - Tone in marketing: confident and enthusiastic — "Build something extraordinary." - Tone in error state: confident and calm — "We have identified the issue. Here is what we are doing about it." - Tone in crisis: confident and transparent — "We take this seriously. Here are the facts."

The voice attribute (confident) persists. The tone (enthusiastic, calm, transparent) adapts.


DEFINING BRAND VOICE

Step 1: Voice Attributes (3-5 Adjectives)

RULE: define exactly 3-5 adjectives that describe how the brand communicates. Not what it does — how it sounds.

PROCESS: 1. Gather stakeholders (founder, marketing, customer-facing staff). 2. Each person writes 10 adjectives that describe how the brand SHOULD sound. 3. Cluster similar words. Vote on the top 5. 4. Refine: replace vague words ("professional") with specific ones ("precise").

EXAMPLE set: Clear, Confident, Warm, Direct, Purposeful

BAD attributes: "innovative" (describes product, not voice), "professional" (too vague), "unique" (meaningless).

Step 2: "We Are / We Are Not" Pairs

RULE: every voice attribute needs a boundary. Without the "we are not" side, the attribute is useless.

We Are We Are Not
Confident Arrogant
Warm Saccharine or fake
Direct Blunt or rude
Clear Dumbed-down or patronizing
Purposeful Preachy or self-important
Witty Sarcastic
Honest Brutal
Enthusiastic Hyperbolic
Helpful Pushy
Calm Cold or detached

RULE: 5-8 pairs is the sweet spot. Fewer than 5 lacks precision. More than 8 is unmanageable.

Step 3: Voice Do's and Don'ts Table

RULE: abstract attributes only become actionable through concrete examples.

Attribute Do Don't Example (Do) Example (Don't)
Clear Use short sentences. One idea per sentence. Use jargon, nested clauses, or hedging language. "Your trial ends in 3 days." "Please be advised that your complimentary trial period will be reaching its conclusion in approximately 3 business days."
Confident State things directly. Use active voice. Use "maybe", "we think", "possibly", or passive constructions. "This fixes the issue." "We believe this should potentially resolve the issue."
Warm Use "you" and "your". Acknowledge the person. Be robotic, impersonal, or form-letter generic. "Thanks for reaching out. Here is what is happening." "Your inquiry has been received and is being processed."
Direct Lead with the action or the answer. Bury the point in preamble. "To change your plan, go to Settings > Billing." "There are several options available. Depending on your current subscription tier, you may want to consider..."

MESSAGING HIERARCHY

Structure

A messaging hierarchy organizes what the brand communicates from most abstract (brand promise) to most concrete (proof points).

BRAND PROMISE
  The single, overarching commitment the brand makes.
  One sentence. Unchanging. The north star.
  Example: "Make enterprise-grade software accessible to every business."

VALUE PROPOSITIONS (3-5)
  The key benefits that support the promise.
  Each one answers: "Why should I care?"
  Example: "Built for hyperscale from day one."
  Example: "ISO 27001 compliant without the enterprise price tag."
  Example: "Your own dedicated AI development team."

PROOF POINTS (per value proposition)
  Evidence that the value proposition is true.
  Data, features, case studies, testimonials.
  Example: "Handles 100,000 concurrent users on the same architecture that serves 1."
  Example: "Passed SOC 2 Type II audit in Q4 2025."

REASONS TO BELIEVE
  The underlying credibility — why the audience should trust this brand.
  Track record, technology, team, methodology, partnerships.
  Example: "Built by 54 specialized AI agents, each trained in a single discipline."

Messaging Hierarchy Rules

RULE: the brand promise is one sentence. If it needs a paragraph, it is not sharp enough. RULE: value propositions are benefits, not features. "Hyperscalable architecture" is a feature. "Grows with you from 1 user to 100,000 without migration" is a benefit. RULE: every proof point must be verifiable. No vague claims. RULE: the hierarchy flows downward: promise -> value props -> proof points -> reasons to believe. Marketing uses all four layers. A tweet might only use one. RULE: update the proof points quarterly as new evidence emerges. The higher layers change rarely.


MESSAGE ARCHITECTURE (BLOOMSTEIN METHOD)

What It Is

A message architecture (Margot Bloomstein's framework) is a prioritized set of brand attributes — terms, phrases, or statements — arranged hierarchically to convey messaging priorities and communication goals.

It differs from a messaging hierarchy in focus: a messaging hierarchy defines WHAT to say. A message architecture defines HOW it should feel.

The BrandSort Process

  1. Compile adjectives. Prepare 80-100 cards, each with a brand attribute word (e.g., "trustworthy", "innovative", "bold", "approachable", "precise", "playful").
  2. Stakeholder card sort. Each stakeholder sorts cards into three piles:
  3. "Who we are now"
  4. "Who we want to be"
  5. "Never us"
  6. Cluster and prioritize. Group the "who we want to be" cards into 3-5 themes. Rank the themes by importance.
  7. Write the architecture. Formalize the top themes as a prioritized list, with the most important attribute at the top.

Example Message Architecture

1. TRUSTWORTHY     — security, reliability, transparency, competence
2. EMPOWERING      — enabling, educational, supportive, tool-giving
3. PRECISE         — accurate, detail-oriented, data-driven, no fluff
4. APPROACHABLE    — warm, plain-language, human, non-intimidating
5. FORWARD-LOOKING — innovative, modern, progressive, future-ready

RULE: the hierarchy IS the priority. When trustworthy and approachable conflict (e.g., disclosing a security incident), trustworthy wins because it ranks higher. RULE: "never us" cards are as important as "who we want to be" — they define the boundaries. RULE: revisit the architecture annually or after a major brand pivot.

How the Message Architecture Guides Work

Decision How the Architecture Guides It
Homepage headline Must reflect #1 attribute (trustworthy) before any other
Social media voice Lean into #4 (approachable) but never at the expense of #1
Error messages #1 (trustworthy) + #4 (approachable) — honest, warm, clear
Sales copy #2 (empowering) + #3 (precise) — show what they gain with specifics
Crisis communication #1 (trustworthy) dominates — transparency, facts, accountability
Blog content #2 (empowering) + #5 (forward-looking) — teach, inspire

VOICE ACROSS CHANNELS

The Principle

RULE: voice stays consistent. Tone adapts. Formality level adapts. But the underlying personality is recognizable everywhere.

Channel-Specific Adaptations

Channel Formality Length Personality Expression Typical Tone
Website (homepage) Medium Concise High — this is the brand's front door Confident, benefit-focused
Website (docs/help) Low-medium As long as needed Low — utility first Clear, precise, helpful
Email (marketing) Medium Short-medium Medium-high Enthusiastic, personal
Email (transactional) Medium Brief Low Factual, clear
Social — LinkedIn Medium-high Medium Medium Professional, thoughtful
Social — Instagram Low Very short High Visual-first, casual, human
Social — X/Twitter Low Ultra-short Medium-high Punchy, conversational
Blog posts Medium Long-form Medium Educational, authoritative
In-app UI copy Low Minimal Low Functional, clear (see ux-writing.md)
Customer support Medium As needed Medium Warm, empathetic, solution-focused
Press releases High Structured Low Formal, factual
Investor comms High Structured Low Precise, data-driven, confident

RULE: the same news (e.g., a new feature launch) sounds different on LinkedIn than on Instagram. The facts are the same. The delivery adapts.

EXAMPLE — same announcement, different channels: - LinkedIn: "We are launching real-time collaboration. Here is why it matters for teams managing complex projects, and the engineering decisions behind it." - Instagram: "Real-time collab is here. Build together, ship faster." [with visual] - Email: "New: real-time collaboration. Your team can now edit together in real time. Here is how to enable it." - In-app: "New: Real-time collaboration. Enable it in Settings > Team." [banner]


BUILDING A VOICE DOCUMENT THAT WORKS

Why Most Voice Guides Fail

Most voice guides fail because they are: 1. Too abstract — "Be authentic" means nothing without examples. 2. Too long — a 40-page PDF that nobody opens. 3. Written once, never updated. 4. Not integrated into the workflow — they live in a Google Drive folder, not in the tools people use.

The One-Page Voice Chart

RULE: the core voice guide fits on one page. Extended documentation can exist, but the essentials are one page.

VOICE CHART TEMPLATE
====================

BRAND: [Name]
VOICE IN THREE WORDS: [Adjective], [Adjective], [Adjective]

WE ARE:                    WE ARE NOT:
- [attribute]              - [boundary]
- [attribute]              - [boundary]
- [attribute]              - [boundary]
- [attribute]              - [boundary]
- [attribute]              - [boundary]

WORD PREFERENCES:
  Use          |  Instead of
  -------------|-------------
  [preferred]  |  [avoided]
  [preferred]  |  [avoided]
  [preferred]  |  [avoided]
  [preferred]  |  [avoided]
  [preferred]  |  [avoided]

SAMPLE COPY:
  Success:  "[example]"
  Error:    "[example]"
  CTA:      "[example]"
  Welcome:  "[example]"
  Support:  "[example]"

Extended Voice Documentation

Beyond the one-page chart, a complete voice system includes:

  1. Voice Chart (1 page) — the quick reference. Printed, pinned, shared.
  2. Tone Matrix (1-2 pages) — maps tone shifts per scenario and severity.
  3. Word List (ongoing) — preferred terms, banned terms, with rationale.
  4. Copy Examples Library (ongoing) — real approved copy organized by scenario type.
  5. Channel Adaptation Guide (1 page) — how voice flexes per channel.
  6. Review Checklist — 10-item checklist for any copy review (see below).

Voice Review Checklist

  • [ ] Does this sound like our brand? Read it aloud — would a user recognize it as us?
  • [ ] Does it match the message architecture priority? (Attribute #1 should be most evident.)
  • [ ] Is the tone appropriate for the context? (Error copy should not sound like marketing.)
  • [ ] Are we using preferred terms from the word list?
  • [ ] Is it as short as it can be without losing meaning?
  • [ ] Does it lead with the benefit or the action, not background?
  • [ ] Is it free of jargon, hedging language, and corporate speak?
  • [ ] Does it work for the channel it will appear on?
  • [ ] Would a new user understand it without prior context?
  • [ ] Has it been tested aloud? Does it sound human?

ANTI-PATTERNS

ANTI_PATTERN: defining voice as a list of values ("integrity, innovation, excellence") FIX: values describe what the company believes. Voice describes how it speaks. A company that values "innovation" can still sound boring if the voice is not defined separately.

ANTI_PATTERN: voice guide that is only positive ("We are friendly, warm, helpful") FIX: include the "We are not" boundary. "Warm but not saccharine." Without boundaries, warm drifts into twee.

ANTI_PATTERN: copying another brand's voice wholesale FIX: voice must be distinctive. If you sound exactly like Mailchimp, you are not Mailchimp — you are a copy. Define what makes YOUR voice different.

ANTI_PATTERN: having one voice for marketing and a completely different one for product/UI FIX: the voice is the same. The tone and formality adapt. A user who encounters marketing and then the product should feel continuity, not whiplash.

ANTI_PATTERN: voice guide written by an agency, handed off, and never internalized by the team FIX: the team that writes the copy must own the voice guide. Workshop it together. Update it together. If an agency builds it, the team must co-author it.

ANTI_PATTERN: no voice guide at all — "we just know how we sound" FIX: what happens when the team doubles? When a new copywriter joins? When an AI agent writes on your behalf? Implicit voice does not scale. Document it.


VOICE IN AN AGENTIC CONTEXT

CONTEXT: GE's agents (including Rick) write copy on behalf of client brands. This creates a unique challenge: the agent must suppress its own "voice" and adopt the client's voice.

RULE: every client project begins with a voice definition exercise. No copy is written until voice is documented. RULE: the client's voice chart is stored in the project wiki and injected into every agent's context when working on that client. RULE: voice consistency is tested by having a different agent (not the original writer) review copy against the voice chart. RULE: when the client has no existing voice guide, Rick leads the BrandSort exercise and produces the voice chart as a deliverable BEFORE any campaign work begins. RULE: agents must NEVER inject GE's own voice into client work. GE's voice is for GE communications only.


REFERENCES

  • Bloomstein, Margot. Content Strategy at Work. Morgan Kaufmann, 2012.
  • Bloomstein, Margot. Trustworthy. Page Two, 2021.
  • Flaum, Sandra and Flaum, Jonah. "The 100-Day Communications Plan." (messaging hierarchy model)
  • Nielsen Norman Group. "The Four Dimensions of Tone of Voice." nngroup.com.
  • Mailchimp Content Style Guide: styleguide.mailchimp.com (widely referenced industry benchmark).
  • Shopify Polaris Content Guidelines: polaris.shopify.com/content.
  • Google Material Design Writing: m3.material.io/foundations/content-design/overview.
  • Erika Hall. Conversational Design. A Book Apart, 2018.