How to Build a Brand Guide

Most brand guides are written to impress and then never opened again: sixty pages of mission poetry, mood boards and logo autopsies that no one consults on a Tuesday afternoon while writing a customer email. The brand guide a small company actually needs answers concrete questions in seconds and is used weekly — by your team, your freelancers, and increasingly by your AI tools. This guide walks you through building exactly that: six working sections plus the governance that keeps them alive, each with the decisions behind it and a worked example for a fictional logistics startup called Freightline. At the end, you'll turn the whole thing into a machine-readable brand kit that AI agents load before they write or design a single thing in your name.

Guide: https://blackcard.dev/guides/how-to-build-a-brand-guide
Markdown version (fetch this for the full method): https://blackcard.dev/guides/how-to-build-a-brand-guide.md

I want to build a working brand guide for my company using the method from BlackCard's guide "How to Build a Brand Guide": positioning (competitive-alternative framing, three tests), voice spec (rules, word lists, sounds-like pairs, per-channel registers), logo system (variants, clear-space formula, minimum sizes, don'ts), color (palette from one brand color, 60/30/10, functional colors, contrast-checked pairs), typography (two typefaces chosen by job, full scale), the five application templates, governance, and a machine-readable brand kit for AI tools.

Interview me first, one question at a time, and keep it concrete:
1. What does my company do, for whom — and what would those customers do if we didn't exist (the real competing alternative)?
2. How do we talk today? Ask me to paste 3-4 real texts (emails, website copy, posts) and tell me what voice they already have.
3. What assets exist: logo files and variants, any colors as hex values, fonts in use, where they all live.
4. Who must be able to use the guide without asking me: team, freelancers, agencies, AI tools?
5. What are the five assets we produce most (emails, posts, slides, proposals, pages)?

Then produce: (a) a draft positioning paragraph in my words framed against the real alternative, tested against the three tests (a stranger could guess our website's tone from it; it excludes someone; no adjective we couldn't defend); (b) a first-pass voice spec derived from my pasted texts; (c) the gap list — every decision still missing before the remaining sections can be written, each with 2-3 options and a recommendation. Challenge vague answers — "quality and innovation" is not a positioning.

Why a working guide beats a brand book

A brand guide has one job: letting someone who is not you produce material that looks and sounds like your company, without asking you. Every page that doesn't serve that job dilutes the pages that do.

The sixty-page version fails in a specific way: the rules that matter (which blue? how do we open emails? can the logo sit on a photo?) are buried between brand-essence pyramids and archetype essays, so people stop looking things up and start guessing. Six months later you have four shades of blue in circulation and a LinkedIn presence that sounds like three different companies — exactly what the document was supposed to prevent.

The working version is six sections plus a governance page. Write it in a focused week, use it for years. And since your AI tools now produce a growing share of your drafts, the guide doubles as their standing instructions — a brand guide that only humans can parse is now half a brand guide.

Section 1: Positioning — one paragraph, framed against the alternative

Positioning is not a slogan; it is the decision everything else inherits from. The frame that works: position against the competitive alternative — not your favorite rival, but what your customers would actually do without you. Often that isn't a competitor at all; it's a spreadsheet, an agency, an intern, or doing nothing. Your paragraph must make a specific person prefer you over that.

The template: who we serve (narrow enough to exclude someone) + what we do for them (outcome, not activity) + why us over the alternative (one defensible difference).

Freightline, filled in:

Freightline books same-week freight for small manufacturers in the DACH
region who are too small for enterprise logistics contracts. We handle the
carrier search, paperwork and tracking through one dashboard, so a two-person
operations team ships like a corporation. Unlike freight marketplaces —
where you are one request among thousands — a named human checks every
booking before it goes out.

A positioning paragraph must pass three tests before it goes in the guide:

1. The stranger test. Someone who has never heard of you reads it and correctly guesses what your website, sales deck and support emails should feel like. If they can't, it's too abstract.
2. The exclusion test. It visibly leaves someone out (here: enterprises with logistics departments, companies outside DACH). A positioning that excludes no one convinces no one.
3. The courtroom test. No adjective you couldn't defend with evidence. "Innovative", "best-in-class" and "passionate" lose in court; "a named human checks every booking" is a checkable fact.

Everything that follows — voice, color, type, templates — gets judged against this paragraph. When a design review stalls, the question is never "do we like it?" but "does it serve the paragraph?"

Section 2: Voice — derived, not decorated

Voice is the section your team uses most and the one that feeds your AI tools directly. Don't invent it from adjectives; derive it from the positioning. Freightline serves busy two-person ops teams and differentiates on human accountability — so its voice must be fast to read, concrete, and calm about problems (accountable people don't hide behind passive voice). Each voice rule should trace back to the paragraph like that.

The full spec — rules, word lists, example pairs, and the register per channel:

FREIGHTLINE VOICE SPEC  v1.2

RULES (each traces to the positioning)
1. Plain words over logistics jargon — our buyer is an ops person, not a
   freight forwarder. If a customer wouldn't say it, we don't.
2. Short sentences, one idea each. Busy people writing to busy people.
3. Concrete over abstract: "booked by Thursday", never "fast turnaround".
4. We say "we" and "you". Never "the user", never passive voice on bad news:
   "we missed the pickup", not "the pickup was missed".
5. Calm, never salesy. We state, we don't shout. No exclamation marks in
   product copy; at most one per marketing piece.

WORDS WE USE            WORDS WE NEVER USE
booking                 solution
carrier                 synergy / streamline
on time / late          best-in-class
price                   competitive pricing
problem                 challenge (when we mean problem)
check / checked         leverage
dashboard               platform (as a noun for us)

SOUNDS LIKE US / NOT US
1. "Your pallet leaves Tuesday. Track it here."
   NOT: "Your shipment has been scheduled for expedited processing!"
2. "We missed the pickup window. Here is what we're doing about it."
   NOT: "Due to unforeseen circumstances, minor delays may occur."
3. "Three carriers quoted. The middle one is the best deal — here's why."
   NOT: "Leverage our network of 500+ certified logistics partners!"
4. "This route got 11% cheaper since March. Want us to rebook?"
   NOT: "Unlock significant savings opportunities today!"
5. "You're paying for speed you don't use. The slower option saves 240 EUR."
   NOT: "Our flexible solutions adapt to your unique needs."

REGISTER PER CHANNEL  (same voice, different volume)
Website:      full sentences, benefit-first, one idea per section.
Sales email:  shorter still; first line states why we're writing;
              one ask per email; no warm-up paragraphs.
Social:       observations and numbers from real operations; it may have
              a point of view; never engagement-bait questions.
Product UI:   imperative labels ("Book carrier"), no humor in errors.
Error msgs:   what happened + what we're doing + what you can do.
              Apologize once, concretely. Never "oops".

The sounds-like/not-us pairs do more work than any adjective — when in doubt, add another pair rather than another adjective. And the register table is what stops "one voice" from meaning "one volume everywhere": your error messages and your LinkedIn posts share DNA, not a template.

Section 3: The logo system

Not "the logo" — the system: which mark exists for which job, plus rules measured in formulas rather than vibes.

• The variants. A primary logo (full wordmark, your default), a secondary/stacked version for narrow spaces, and a standalone mark (icon) for avatars, favicons and tiny placements. If you only have the wordmark today, that's fine — say so in the guide and define where it simply doesn't go, instead of letting people squeeze it into a 32px circle.
• Clear space as a formula. Define it relative to the logo itself so it scales: keep a margin equal to the height of the tallest letter (the "F-height") on all sides. Nothing enters that space — no text, no badges, no canvas edge. A formula survives resizing; "some breathing room" does not.
• Minimum sizes, screen and print. Below legibility, the logo damages recognition instead of building it. Set both numbers: e.g. wordmark at least 24px tall on screen / 10mm in print; below that, use the mark alone; below 16px / 6mm, use nothing.
• Backgrounds. One light version, one dark version. The logo sits on white, on your primary dark, or on photographs only behind a scrim (a 40–60% dark overlay). No other recolors — ever.
• The don'ts — each with its why. Don't stretch (breaks the letterforms people recognize faster than they read), don't rotate (reads as error, not energy), don't recolor (color is half of recognition at a distance), don't add shadows/outlines/gradients (they fight small sizes and dark mode), don't place on busy photos without a scrim (contrast collapse), don't redraw it "quickly" in a slide (now two logos exist, and the wrong one multiplies). In the actual guide, show each don't as a crossed-out picture; people copy pictures, not prose.

If your logo needs more rules than this, the problem is usually the logo.

Section 4: Color — a palette built from one color, with rules

Most small companies have one real brand color. Build the palette from it instead of collecting favorites:

1. Primary: your brand color, plus one dark neutral derived from it (shift your brand hue toward near-black instead of using pure #000 — it keeps dark surfaces feeling owned).
2. Accent: one — exactly one — high-energy color reserved for the primary action. Its scarcity is its power.
3. Neutrals: a warm or cool white for backgrounds and a mid gray for secondary text, both leaning toward the brand hue.
4. Functional: green/red/yellow for success/error/warning, desaturated or hue-shifted until they sit comfortably next to the palette (a pure #FF0000 error will always look like a foreign object next to a warm palette).

Then the part everyone skips — usage rules and verified contrast:

FREIGHTLINE COLOR  v1.2

PRIMARY    Freight Navy   #122B44   headers, primary surfaces, logo
ACCENT     Signal Orange  #FF6A2B   the one clickable thing per screen
NEUTRALS   Paper          #F7F6F3   backgrounds
           Slate          #5C6670   secondary text, borders, icons
FUNCTIONAL Green          #1F9D55   success only
           Red            #D64545   errors + destructive only
           Yellow         #E8B931   warnings only

USAGE — 60 / 30 / 10
~60% Paper (space), ~30% Navy (structure), ~10% Orange (action).
If a screen feels off, count: it is almost always too much 10.

RULES
- Signal Orange appears once per screen/asset: on the thing we want
  clicked. Two orange things = no primary action.
- Functional colors never decorate. Green means "it worked", not "nature".
- No gradients; no transparencies of brand colors except 8% Navy
  for table stripes.

CONTRAST-CHECKED TEXT PAIRS (WCAG AA)
  Navy  #122B44  on Paper #F7F6F3   13.9:1  body text         PASS
  Paper #F7F6F3  on Navy  #122B44   13.9:1  inverse text      PASS
  Slate #5C6670  on Paper #F7F6F3    5.1:1  secondary >=16px  PASS
  Paper #F7F6F3  on Orange #FF6A2B   2.9:1  NOT for text — large
                                            headlines/icons only
  Navy  #122B44  on Orange #FF6A2B   4.8:1  button label      PASS

That last block is the difference between a palette and a system: every text/background combination anyone will reach for, pre-checked, with the failing one explicitly forbidden. (This is also the first thing a UI/UX audit flags when it's missing.)

Section 5: Typography — two typefaces, chosen by their job

Choose type by job, not by mood board. The headline face carries personality at large sizes — it may be characterful because it appears in short bursts. The text face carries information at 14–16px for paragraphs at a time — it must be boring in the best way: large x-height, open counters, good at small sizes, with real weights. One family that does both jobs (a solid variable sans) is even better than two — fewer licenses, fewer failure modes.

Then the scale, written down completely so documents stop being font lotteries:

FREIGHTLINE TYPE  v1.2

FACES   Headlines: Space Grotesk (Semibold only)
        Everything else: Inter

SCALE            size/line     weight      where
display          48/52         Semibold    hero only, max 2 lines
h1               32/40         Semibold    page titles
h2               24/32         Semibold    section titles
h3               18/26         Semibold    card titles
body             16/24         Regular     paragraphs, default
secondary        14/20         Regular     meta text, in Slate
label            13/16 CAPS    Medium      +2% letter-spacing
Nothing below 13px, anywhere, ever.

RULES
- Body is never bolded for emphasis; rewrite the sentence instead.
- Italics only for titles of works.
- Line length 60-75 characters for body text.

FALLBACK STACKS (where custom fonts fail)
web/app:   Inter, -apple-system, "Segoe UI", Roboto, sans-serif
email:     Arial, Helvetica, sans-serif
docs:      system sans (Calibri/Helvetica) — never substitute a serif

Name the fallbacks explicitly so "it looked different in Outlook" stops being a surprise, and license both faces for web, desktop and email use on day one — retro-licensing after a trademark letter is the expensive version.

Section 6: The application layer — five templates that end arguments

Rules become a brand only when they're applied, and the fastest way to get consistent application is to specify the five assets your company produces most — precisely enough that a freelancer gets them right without asking:

1. Email signature. Name (body, Semibold), role and company (secondary, Slate), one phone, one link. Logo mark at 20px, left-aligned. No banners, no quotes, no inspirational PS.
2. Slide deck — title and content masters. Title slide: Navy background, Paper display type, logo bottom-left at fixed size. Content slide: Paper background, h2 title, max 5 bullets of body, one orange element maximum. If the number or chart is the point, the slide shows the number — not three paragraphs.
3. Social post anatomy. Square or 4:5. Paper or Navy background, one h2-sized statement (max 12 words), mark top-right at fixed size, footer with handle in label style. Photos always behind the scrim when text sits on them. Caption follows the social register: observation + substance, no bait.
4. Invoice / proposal header. Logo top-left at minimum-size-plus, document title as h1, meta block (number, date, client) as labels + secondary. Navy rule line under the header, nothing else decorated — money documents are where calm reads as competence.
5. Website hero. One display-size statement serving the positioning paragraph, one orange CTA, one supporting visual. If the hero needs a second CTA, the page is two pages.

Each template lives in the guide as a picture with a one-line caption naming the rules at work ("Navy header, one orange CTA, Inter body, clear space respected"). Examples end arguments: when a new freelancer asks "like this?", the answer is a page number instead of a meeting.

Governance: the part that keeps it alive

Three habits and one role separate a working brand guide from a dead PDF:

• One source of truth. The guide lives at exactly one link (web page, Notion, Figma — anything linkable and updatable). Copies circulating as attachments are how versions fork and the guide dies. Export PDFs for external partners; the link stays canonical.
• A named owner. One person approves changes and answers "is this on-brand?". Not a committee — a person, named in the guide itself. Their job is small: minutes per week, decisive when it matters.
• Changes by precedent, with a changelog. When reality isn't covered (first podcast cover, first trade-show booth), the owner decides once, the decision enters the guide, the changelog gets one line with a date and version bump (v1.2 → v1.3). The guide grows from six pages to maybe ten over years — all earned, none poetry.
• The quarterly drift review. Fifteen minutes, four questions: pull the last 10 public assets — do they pass the squint test as one company? Any new colors/fonts in circulation that aren't in the guide (kill or adopt)? Which rule was asked about most (clarify it)? Does the positioning paragraph still pass its three tests (markets move)? Log the answers; two drifting quarters in a row means the guide has stopped being used, and that's an ownership problem, not a design problem.

The machine-readable brand kit

The newest reason to write this document well: it is no longer only for humans. A growing share of your drafts — posts, replies, ad variants, page copy — is produced by AI tools, and they follow whatever instructions they're given. A brand guide that lives in a PDF is invisible to them. So the final step is converting the decisions above into a structured brand-kit file your agents load before every run:

# freightline-brand-kit.yaml  v1.2  owner: brand@freightline.example
positioning: >
  Freightline books same-week freight for small DACH manufacturers too
  small for enterprise logistics contracts. One dashboard for carrier
  search, paperwork, tracking. Unlike marketplaces, a named human checks
  every booking.
voice:
  rules:
    - plain words over logistics jargon
    - short sentences, one idea each
    - concrete over abstract ("booked by Thursday")
    - "we" and "you"; active voice on bad news
    - calm, never salesy; no exclamation marks in product copy
  words_use: [booking, carrier, on time, late, price, problem, checked]
  words_never: [solution, synergy, best-in-class, leverage, streamline]
  sounds_like: "Your pallet leaves Tuesday. Track it here."
  not_us: "Your shipment has been scheduled for expedited processing!"
  register: {social: "observation + numbers, no bait",
             email: "first line says why; one ask",
             error: "what happened + what we do + what you can do"}
visual:
  palette: {navy: "#122B44", orange: "#FF6A2B", paper: "#F7F6F3",
            slate: "#5C6670", green: "#1F9D55", red: "#D64545"}
  rules: ["orange exactly once per asset", "60/30/10 paper/navy/orange",
          "functional colors never decorate"]
  type: {headline: "Space Grotesk Semibold", text: "Inter",
         min_size_px: 13}
  logo: {clear_space: "1x F-height", min_px: 24, on_photo: "scrim 40-60%"}
post_anatomy:
  formats: ["1:1", "4:5"]
  statement_max_words: 12
  mark_position: top-right
compliance:
  never_claim: ["guaranteed delivery times", "cheapest on the market"]
  disclosure: "ad posts marked per platform rules"

This file is the voice-and-visual contract for every agent that writes or designs in your name — paste it into your AI assistant's project instructions, and hand it to the marketing agents from our guide on how to run AI agents for marketing, where it becomes the brand system the whole pipeline enforces. One document, versioned with the guide, enforced by every tool.

Guide: https://blackcard.dev/guides/how-to-build-a-brand-guide
Markdown version (fetch this for the full method): https://blackcard.dev/guides/how-to-build-a-brand-guide.md

Act as a brand strategist and designer. Build my company's complete working brand guide following BlackCard's guide "How to Build a Brand Guide": six sections plus governance plus a machine-readable brand kit. A usable document, not a brand book.

Company: [name, what it does, for whom]
The real alternative customers compare us to: [competitor, agency, spreadsheet, doing nothing]
Real texts in our current voice (paste 3-4): [emails, website copy, posts]
Existing assets: [logo files/variants, colors as hex, fonts, where they live]
Our five weekly assets: [e.g. customer emails, LinkedIn posts, slides, proposals, landing pages]

Produce, in this order:
1. Positioning — one paragraph against the real alternative; verify it passes the stranger test, the exclusion test and the courtroom test, and say how.
2. Voice spec — 5 rules each traced to the positioning, words-we-use / words-we-never-use lists, 5 sounds-like/not-us pairs derived from my pasted texts, and a register per channel (website, sales email, social, product UI, error messages).
3. Logo system — variants and their jobs, clear space as a formula, minimum sizes for screen and print, backgrounds, and a don'ts list with the WHY for each.
4. Color — palette built from my brand color (primary, one accent, neutrals, functional), 60/30/10 usage, rules, and a contrast table of every text/background pair with pass/fail at WCAG AA 4.5:1.
5. Typography — two faces chosen by job (or one variable family), the full scale with sizes, line-heights and weights, usage rules, fallback stacks for web/email/docs.
6. Application templates — precise specs for each of my five weekly assets, executable by a freelancer without questions.
7. Governance — the owner's role, change-by-precedent process with changelog and versioning, and the quarterly drift-review checklist.
8. The machine-readable brand kit — a YAML file version of all of the above for AI tools, including compliance lines (claims we never make).
Where information is missing, give me a decision with 2-3 options and a recommendation instead of inventing. Challenge anything generic.

Building brand foundations like this — and the visual identity on top of them — is what our branding work at BlackCard is about. If you want a brand guide your team and your AI tools actually use, book a free call and bring the five assets you produce most.