Brand voice is the most discussed and least understood asset a company has. It gets treated as a document, produced during a rebrand, admired briefly, and then ignored by everyone who actually writes the posts.
The reason it fails is usually that the voice was defined in adjectives. Approachable. Confident. Authentic. Every brand claims these, no brand is differentiated by them, and none of them tell a writer what to actually type on a Tuesday morning with a post due at noon.
A voice becomes usable the moment it is expressed as decisions a person can make while writing.
Instead of "we are approachable," say "we use contractions, we never use the word solution, and we address the reader as you." Instead of "we are confident," say "we make claims without hedging, we do not use the word maybe, and we never apologize for having an opinion." Instead of "we are witty," say "we make one joke per post at most, the joke is never about the customer, and we never explain it."
The difference is that the second version can be followed. The first version can only be interpreted, and every person interprets it differently, which is precisely why brand voice collapses the moment more than one person is writing.
The test is whether two different writers, given the same rules and the same brief, produce work that sounds like it came from the same organization. If they do not, the voice is not defined yet. Run this test before you sign off on any voice document, because it will expose immediately whether you have written a set of rules or a set of hopes.
Vocabulary. The words you use and the words you refuse. This is the fastest lever available and the most neglected.
A short banned-words list does more for consistency than any amount of philosophy. Ours for one client includes solution, leverage, journey, empower, seamless, and elevate, none of which mean anything and all of which appear in every competitor's copy. Removing them alone forced the writing into specificity, because once you cannot say "we empower creators," you have to say what you actually do.
The permitted list matters too. Certain words become yours through repetition, and a brand that consistently uses a small set of distinctive terms builds a vocabulary that people recognize.
Sentence rhythm. Long and considered, or short and direct? Do you use fragments? Do you start sentences with conjunctions? Do you use rhetorical questions, and if so, how often?
These decisions are audible even when readers cannot name them, and they are the single most identifiable feature of a voice. Two brands using identical vocabulary with different rhythms will not sound alike.
Register, and how it shifts. How formal are you, and does that change by context? A voice that is identical in a customer service reply and a product launch is probably too rigid. One that is unrecognizable across them is too loose.
The useful frame is that you are the same person at a dinner party and a job interview. Your character does not change. Your register does. Specify which is which.
Humor. The most dangerous variable, because it is the one most likely to be executed badly by someone imitating rather than possessing it.
If humor is part of your voice, be specific about what kind, because dry, absurd, self-deprecating, observational, and mean are entirely different postures and they are not interchangeable. A brand that is funny in a dry way and hires someone who is funny in an absurd way will produce content that is technically on-brand and completely wrong.
What you never do. Often more useful than what you do.
We do not use exclamation points. We do not jump on trends that have nothing to do with us. We do not reply to criticism with a joke. We do not use the word community to describe our customers. Constraints define character more sharply than aspirations, and they are far easier to enforce.
Consistency breaks in four predictable places, and each has a specific fix.
It breaks when new people join. The onboarding artifact matters far more than the strategy deck, and almost nobody builds one.
Give a new writer twenty examples of posts that are correct and ten that are wrong, with one sentence explaining each. That teaches faster than any set of principles, because voice is learned by pattern recognition rather than by rule memorization. A person who has read thirty annotated examples will write better on-brand copy than one who has read a forty-page guidelines document.
It breaks across platforms. A voice has to flex without breaking, and the failure mode in both directions is common. Some brands post identical copy everywhere, which reads as lazy on the platforms where it does not fit. Others become an entirely different entity on each platform, which means they have no voice at all.
The rule we use is that the character stays fixed and the register adapts. Your opinions, your vocabulary, and your sense of humor travel. Your formality and your length do not.
It breaks under pressure. Particularly during a crisis, when panicking executives revert to corporate language and the voice evaporates precisely when the audience is paying most attention.
This is the most damaging break, because a brand that sounds human for two years and then sounds like a legal department the moment something goes wrong has just told its audience which one was the performance. Decide in advance what you sound like when something goes wrong, and write two examples. You will not be able to invent it in the moment.
It breaks through committee. Every additional approver pulls the writing toward the safe middle.
The mechanism is that each reviewer removes the thing that makes them personally uncomfortable, and the aggregate of those removals is a piece of writing with nothing in it. If your voice is genuinely distinctive, it will make someone in the approval chain uneasy, and the willingness to publish anyway is the entire difference between brands with a voice and brands with a style guide.
A real voice costs you something. It will alienate people, it will draw criticism, and someone senior will eventually ask whether it is a good idea to sound like this.
That discomfort is not a sign the voice is wrong. It is the price of being distinguishable. The alternative, being pleasant and unmemorable, is not actually safer. It just fails more quietly, and by the time anyone notices, the failure is attributed to the market rather than to the decision to sound like everyone else.