# Brand Systems

Canonical: https://andydataguy.com/wiki/attention-engineering/default/brand-systems

Author: Anand Houston (AndyDataGuy)

The needle snaps from one setting to the next and never rests between them. A piece of writing works the same way: it's locked into one register, one mode of writing, from its first sentence. The bars redraw at every position because sentence length belongs to the register itself.

Most brand voice work ships as a tone-of-voice document: forty pages, written by a strategist, read once at onboarding and then forgotten. Ten months later, three different writers are producing the brand's pages and you can tell which writer wrote which page from across the room. Brand voice failed because nobody designed it as something a different person could ship and someone could later check.

Brand systems is the alternative: voice gets decomposed into three load-bearing components, each one with explicit rules and an audit method. The [Voice Fingerprint work](/case-studies/voice-fingerprint-deliverable) made that audit measurable across hundreds of source assets before the campaign copy was rebuilt. The components compose so the same voice survives across channels, across writers, and across years. The system is the rules plus the audit, and the document is just where the rules are written down.

## Three components · register, vocabulary, mythology

Tone is the output of a brand voice, the part the reader experiences. The input is three components that produce that tone consistently. Treat tone as the output and you can engineer it. Treat tone as the input and you're stuck describing it with adjectives.

Register. A register is the mode the writing is operating in. My site has three: DEFAULT (the canonical wiki entry, structured, thorough), CASUAL (the homepage, the field note, the chat window, conversational with permission), and PUBLIC (press material, the cold-traffic landing page, more reserved). Each register has its own grammar, sentence-length distribution, and emotional bandwidth. A piece is locked into one register from the first sentence and doesn't bounce between them. The reader knows within ten seconds which register they're in, and the experience makes sense.

Vocabulary. Vocabulary is the list of words and phrases that are explicitly allowed, banned, or required for certain concepts. The list is short and does a lot of work. My banned list includes em dashes, the not X. it's Y pattern, performative announcements, and a roster of AI-tell vocabulary (delve, leverage, utilize, holistic, streamline, empower, transform, robust, comprehensive). The banned list is the part most operators skip and the part that does most of the work. Every banned phrase is a category of AI-shaped sentence that a pre-commit hook can detect and refuse to ship.

Tone is what the reader experiences, so it's an output nobody can set directly. Register, vocabulary and mythology are the inputs, and each one is specifiable, teachable to a new writer, and checkable afterwards.

Mythology. Mythology is the set of recurring stories, archetypes, and named concepts that appear across pieces. Mine include the Beggar and the Chooser, the Velvet Rope, the Scar Tissue Audit (the four diagnostic questions I ask on a first call), the Lexicon of Pain, and People · Product · Process. Each one is a named handle for an idea the body of work returns to repeatedly. The mythology compounds because a reader who has met the Beggar in one piece recognizes the Beggar in the next, and that recognition is what makes the work feel like part of a body instead of a feed of disconnected posts.

## Register design · how to define a mode

Each register lives or dies on whether a different writer can pick it up cold and produce a piece that lands in the right mode. The discipline that makes that possible is the register spec: a page or two per register, written down and treated as canonical.

The spec answers six questions. Who is the reader for this register, in concrete terms? What did they come here for? What is the typical sentence length and how much variance? How much emotional bandwidth is allowed, on a five-point scale from clinical to vulnerable? What are five representative example sentences that demonstrate the mode? What moves are explicitly out of bounds in this register?

The example sentences are the part that carries the weight. Most operators write the abstract description and skip the examples. The abstract description is debatable, and the examples aren't. A new writer reads the five examples and absorbs the pattern in a way no list of adjectives can produce. Andy's CASUAL register example: most agency postmortems blame the funnel, the offer, or the traffic. Almost no one audits the qualification step. That sentence is the register, distilled. It does the descriptive work, names the contrarian frame, and signals permission to push back, all in two clauses. The register is whatever produces sentences that land in this neighborhood.

## Vocabulary control · the banned list does the heavy lifting

The banned list is where most brand voice systems get serious. Adjective lists are easy to write and impossible to enforce, while a script can enforce a banned-token list. The banned list's discipline is converting a vague preference into a checkable rule.

My banned list operates at three levels. For punctuation, it's zero em dashes, period. For patterns, it's zero not X. it's Y constructions, zero here's the thing openers, and zero let me be honest . For vocabulary, it's a roster of AI-tell words that have lost meaning through corporate overuse (delve, crucial, leverage as a verb, utilize, facilitate, holistic, synergy, empower, robust, comprehensive in the marketing sense, streamline, transform as a marketing word). Each entry is enforceable: a pre-commit hook can grep for any of them and refuse the file.

The banned list compounds in a way descriptions don't. Once a writer has internalized the list, they stop generating banned candidates, and the work writes faster because the substitution becomes automatic: leverage our platform never reaches the page because the writer has stopped producing it. Six months in, the list barely needs enforcement because the brain has adopted the constraint.

This is why the banned list carries more weight than any description of the voice. An adjective needs a person to interpret it. A token is checkable by a script at the moment of writing, so the constraint arrives before the sentence ships rather than after.

The required-vocabulary list is the banned list's smaller cousin, and it matters just as much. Some concepts must always be named with their canonical handle instead of a paraphrase. PPP is always People · Product · Process when first introduced. The Velvet Rope stays the Velvet Rope instead of becoming the qualification process , and the Scar Tissue Audit never turns into the customer research method . Required vocabulary is what makes the mythology compound. Without it, every piece reinvents its own terms and the reader never accumulates a shared lexicon.

## The audit method · what you actually grade

The audit method is what turns the brand from a wish into something checkable. It grades three axes, each one pass-fail, with no numeric scores.

The first axis is register fidelity. Read the piece and ask whether it stayed in one register from open to close or bounced. Bouncing is the most common failure mode and the easiest to spot: a CASUAL piece that suddenly shifts into corporate-DEFAULT for a paragraph fails.

The second axis is vocabulary compliance. Run the script, and if it finds any banned token, banned pattern or em dash, the piece fails. The script is automated. The writer should run it before the audit pass, and the auditor should run it again as confirmation. The required-vocabulary check asks whether the piece used canonical handles for the concepts it introduced or paraphrased a named idea into something generic. Paraphrasing a named concept is a fail.

The third axis is mythology presence: did the piece earn its place in the body of work? A piece that doesn't reference any existing mythology, doesn't introduce a new named concept, and doesn't link to any canonical entry exists in isolation. Isolation can be appropriate for a single tactical post, but a body of work composed entirely of isolated pieces fails the body-of-work test. The audit catches drift toward isolation and triggers correction.

The three axes take ten minutes per piece and are repeatable across a team. The audit is the moment the brand stops being something the founder holds in their head and becomes something the team can ship without the founder.

## Where to start

There are three starting points, ordered by difficulty and impact.

Easiest, do today. Write down ten banned words and patterns. Pick the AI-tell vocabulary you most resent finding in your drafts. Write the list down, paste it on the wall, run a grep against your last ten pieces, and count the violations. The number is usually higher than expected. The fix is mechanical from there.

Medium, this week. Name your two or three registers and write the spec for each. The five example sentences per register are the part that carries the weight, and they should be the best ones you've already shipped, not idealized future writing. The spec codifies what you do at your best.

Hardest, this month. Catalog your existing mythology. Walk every piece you've shipped in the last year and list every named concept (Beggar, Velvet Rope, Scar Tissue Audit). Pick the canonical handle for each one. Note any concept that appears under three different names; those are paraphrasing failures. Lock the canonical handle for each, write a one-line definition, and require it across the body of work going forward. The mythology will start compounding within two months because the named concepts will start showing up in audience language.

PRINCIPLE

Voice is a system with three components: register, vocabulary, mythology. Each component has rules, and the rules are auditable. The audit is what makes the brand survive a team. For the broader theory of how attention, trust, narrative, register, and distribution compose, see [Standard Model of Effective Media](/wiki/attention-engineering/default/standard-model-effective-media).

### RELATED ENTRIES

[ATTENTION ENGINEERING · NEAR-BOOK
Standard Model of Effective Media](/wiki/attention-engineering/default/standard-model-effective-media)
[ATTENTION ENGINEERING · ~12 MIN
Content Strategy](/wiki/attention-engineering/default/content-strategy)
[ATTENTION ENGINEERING · ~10 MIN
Video Content Engine](/wiki/attention-engineering/default/video-content)

Source: https://andydataguy.com/wiki/attention-engineering/default/brand-systems