Brand Tone of Voice: how to develop
Voice, personality and tone — why they are always confused
The confusion in terminology is not accidental: the three concepts describe different layers of one system, and most brand books mix them into one paragraph. Brand personality is a set of human traits that the brand 'puts on' (friendly, serious, bold). Brand voice is how this personality constantly expresses itself in writing and speech: vocabulary, sentence length, favorite metaphors. Tone is the contextual modulation of voice: the same voice sounds differently in an error payment letter and in a congratulatory letter.
Mailchimp was one of the first to publicly separate voice and tone in their guidelines, and most SaaS companies have since copied this structure — it is now an industry de facto standard. The personality does not change, the voice hardly changes, and the tone flexibly adapts to the situation.
| Concept | What it describes | Does it change by situation | Example |
|---|---|---|---|
| Personality | Brand character traits | No, it is stable | "Competent and friendly" |
| Voice | A sustainable way to express personality in text | No, it is consistent across all channels | Short sentences, address on 'you', no bureaucratic jargon |
| Tone | Voice modulation per context | Yes, it depends on the situation | Calm tone in case of error, festive in case of success |
Framework instead of three adjectives
Lists of adjectives are subjective: what one author considers "friendly" another may find familiar. There are two frameworks that solve this problem at different levels — strategic and operational — and they should be used together, not instead of each other.
Jennifer Aaker's model describes personality through five dimensions — Sincerity, Excitement, Competence, Sophistication, Ruggedness — each of which breaks down into specific facets (for example, Sincerity includes "honest", "down-to-earth", "wholesome"). This allows for quantitative comparison of brands rather than debating whether a brand is "friendly" or "sincere". Nielsen Norman Group proposed another tool — four dimensions of tone in text: funny vs. serious, formal vs. casual, respectful vs. irreverent, matter-of-fact vs. enthusiastic. This is no longer about brand character, but about specific copywriter decisions in a sentence.
| Framework | Level | What it provides |
|---|---|---|
| Aaker's 5 Dimensions | Strategic (personality) | Positioning of brand character relative to competitors |
| NN Group 4 Dimensions | Operational (tone) | Measurable scales for evaluating specific text and its alignment with voice |
From an adjective to a rule that can be verified
The problem with most guidelines is that they stop at the word "friendly" and don't go further. The working model includes four layers for each trait: the trait itself, behavioral definition, contrast statement, and a real example from the company's content. Without these four elements, the trait remains decorative.
- Trait — "friendly"
- Behavioral definition — address the reader in the informal "you", sentences are shorter than 15 words, no jargon or passive voice
- Contrast statement — "friendly but not flippant: we do not sacrifice expertise for camaraderie"
- Real example — a specific sentence from a real letter or product screen, not a fabricated example "just for show"
It is precisely the fourth layer—real examples—that distinguishes the guidelines of Dropbox, Microsoft, Mailchimp, and Slack from most brandbooks: there they cite actual texts rather than abstract illustrations. And according to the research Morrison (2024), the most effective enforcement voice tool at scale was not a list of principles, but a simple list of prohibited words and constructions with ready-made alternatives—because it is easier for an editor to check against a list than to interpret an adjective.
| Prohibited | Why | Alternative |
|---|---|---|
| «To make a purchase» | Bureaucratic language, violates the «casual» tone of NN dimensions | «To buy» |
| «Dear user» | Too formal for a friendly brand | «Hello» / username |
| «In view of the above» | Destroys voice, adds a bureaucratic tone | Therefore |
One voice, different situations: tone matrix
Error — to think that the voice should sound the same everywhere. The voice as a personality is stable, but the tone must flex: what is appropriate in a congratulatory letter will destroy trust in an error letter. A practical tool — tone matrix by typical product situations, which pre-sets the required register for each context.
| Situation | The right tone |
|---|---|
| Success / achieving the goal | Celebratory — festive, with enthusiasm |
| Error | Calm, helpful — calm, with a clear instruction of what to do |
| Loading / waiting | Reassuring — calming, without irony |
| Sensitive requests (data, payment, deletion) | Respectful — respectful, without jokes |
| Learning / onboarding | Encouraging — supportive, patient |
Such a matrix is closed in the guideline in one table, but saves the team dozens of hours of debates 'can we joke here'.
How to test tone on audience, not on team's taste
The team almost always approves its own tone — this is called groupthink, not validation. A working method is similar to brand positioning testing: audiences are shown not the final version, but several alternatives, and are asked to evaluate them on scales, not to say 'like/dislike'.
- Collect 3-4 variants of the same message, differing only in tone (according to NN Group scales).
- Show respondents the variants without brand context (blind test) and ask them to rate each variant on semantic differential scales: formally-informally, seriously-with humor, respectfully-rudely.
- Ask which variant seems 'more similar to the brand you trust' — without hinting at which brand is being discussed.
- Test the texts in real usage context (error screen, not an isolated sentence) — tone is perceived differently when removed from situation.
- Compare results with target positions according to Aaker's dimensions, which were stated during the strategy phase — does the perception match the intention or not.
This takes longer than asking the marketing director's opinion, but this kind of test is what distinguishes a voice that works for the audience from a voice that simply pleases the company internally.
Audit of the current voice and why it drifts when scaling
Before writing a new guideline, it's fair to take a honest look at what already exists. A voice drift audit is simple to do: collect a sample of texts across all channels — website, emails, social media, support — and evaluate each using the same scales from NN Group and the situation matrix. The variation in ratings within a single brand is the measurable drift.
The voice almost always spreads for the same reason: rules exist in the form of words in a PDF that was read once during onboarding and then forgotten, rather than as tools embedded in the workflow. Many authors, many channels, zero enforcement — that's the drift formula. The problem is solved not by new formulations, but by operational mechanisms: forbidden-words checklists in the text editor, tone matrix as a checklist before publishing, review of real examples rather than abstract principles, and including the same rules in prompts for AI-generated content, which increasingly writes drafts on behalf of people.
Investment in such a system pays off not only aesthetically: according to Frontify, brands with consistent communication earn 23% more than competitors with a misaligned voice — that's the argument to show to the CFO when he asks why another document about text tone is needed.
Voice, personality and tone — why they are always confused
The confusion in terminology is not accidental: the three concepts describe different layers of one system, and most brand books mix them into one paragraph. Brand personality is a set of human traits that the brand 'puts on' (friendly, serious, bold). Brand voice is how this personality constantly expresses itself in writing and speech: vocabulary, sentence length, favorite metaphors. Tone is the contextual modulation of voice: the same voice sounds differently in an error payment letter and in a congratulatory letter.
Mailchimp was one of the first to publicly separate voice and tone in their guidelines, and most SaaS companies have since copied this structure — it is now an industry de facto standard. The personality does not change, the voice hardly changes, and the tone flexibly adapts to the situation.
| Concept | What it describes | Does it change by situation | Example |
|---|---|---|---|
| Personality | Brand character traits | No, it is stable | "Competent and friendly" |
| Voice | A sustainable way to express personality in text | No, it is consistent across all channels | Short sentences, address on 'you', no bureaucratic jargon |
| Tone | Voice modulation per context | Yes, it depends on the situation | Calm tone in case of error, festive in case of success |
Framework instead of three adjectives
Lists of adjectives are subjective: what one author considers "friendly" another may find familiar. There are two frameworks that solve this problem at different levels — strategic and operational — and they should be used together, not instead of each other.
Jennifer Aaker's model describes personality through five dimensions — Sincerity, Excitement, Competence, Sophistication, Ruggedness — each of which breaks down into specific facets (for example, Sincerity includes "honest", "down-to-earth", "wholesome"). This allows for quantitative comparison of brands rather than debating whether a brand is "friendly" or "sincere". Nielsen Norman Group proposed another tool — four dimensions of tone in text: funny vs. serious, formal vs. casual, respectful vs. irreverent, matter-of-fact vs. enthusiastic. This is no longer about brand character, but about specific copywriter decisions in a sentence.
| Framework | Level | What it provides |
|---|---|---|
| Aaker's 5 Dimensions | Strategic (personality) | Positioning of brand character relative to competitors |
| NN Group 4 Dimensions | Operational (tone) | Measurable scales for evaluating specific text and its alignment with voice |
From an adjective to a rule that can be verified
The problem with most guidelines is that they stop at the word "friendly" and don't go further. The working model includes four layers for each trait: the trait itself, behavioral definition, contrast statement, and a real example from the company's content. Without these four elements, the trait remains decorative.
- Trait — "friendly"
- Behavioral definition — address the reader in the informal "you", sentences are shorter than 15 words, no jargon or passive voice
- Contrast statement — "friendly but not flippant: we do not sacrifice expertise for camaraderie"
- Real example — a specific sentence from a real letter or product screen, not a fabricated example "just for show"
It is precisely the fourth layer—real examples—that distinguishes the guidelines of Dropbox, Microsoft, Mailchimp, and Slack from most brandbooks: there they cite actual texts rather than abstract illustrations. And according to the research Morrison (2024), the most effective enforcement voice tool at scale was not a list of principles, but a simple list of prohibited words and constructions with ready-made alternatives—because it is easier for an editor to check against a list than to interpret an adjective.
| Prohibited | Why | Alternative |
|---|---|---|
| «To make a purchase» | Bureaucratic language, violates the «casual» tone of NN dimensions | «To buy» |
| «Dear user» | Too formal for a friendly brand | «Hello» / username |
| «In view of the above» | Destroys voice, adds a bureaucratic tone | Therefore |
One voice, different situations: tone matrix
Error — to think that the voice should sound the same everywhere. The voice as a personality is stable, but the tone must flex: what is appropriate in a congratulatory letter will destroy trust in an error letter. A practical tool — tone matrix by typical product situations, which pre-sets the required register for each context.
| Situation | The right tone |
|---|---|
| Success / achieving the goal | Celebratory — festive, with enthusiasm |
| Error | Calm, helpful — calm, with a clear instruction of what to do |
| Loading / waiting | Reassuring — calming, without irony |
| Sensitive requests (data, payment, deletion) | Respectful — respectful, without jokes |
| Learning / onboarding | Encouraging — supportive, patient |
Such a matrix is closed in the guideline in one table, but saves the team dozens of hours of debates 'can we joke here'.
How to test tone on audience, not on team's taste
The team almost always approves its own tone — this is called groupthink, not validation. A working method is similar to brand positioning testing: audiences are shown not the final version, but several alternatives, and are asked to evaluate them on scales, not to say 'like/dislike'.
- Collect 3-4 variants of the same message, differing only in tone (according to NN Group scales).
- Show respondents the variants without brand context (blind test) and ask them to rate each variant on semantic differential scales: formally-informally, seriously-with humor, respectfully-rudely.
- Ask which variant seems 'more similar to the brand you trust' — without hinting at which brand is being discussed.
- Test the texts in real usage context (error screen, not an isolated sentence) — tone is perceived differently when removed from situation.
- Compare results with target positions according to Aaker's dimensions, which were stated during the strategy phase — does the perception match the intention or not.
This takes longer than asking the marketing director's opinion, but this kind of test is what distinguishes a voice that works for the audience from a voice that simply pleases the company internally.
Audit of the current voice and why it drifts when scaling
Before writing a new guideline, it's fair to take a honest look at what already exists. A voice drift audit is simple to do: collect a sample of texts across all channels — website, emails, social media, support — and evaluate each using the same scales from NN Group and the situation matrix. The variation in ratings within a single brand is the measurable drift.
The voice almost always spreads for the same reason: rules exist in the form of words in a PDF that was read once during onboarding and then forgotten, rather than as tools embedded in the workflow. Many authors, many channels, zero enforcement — that's the drift formula. The problem is solved not by new formulations, but by operational mechanisms: forbidden-words checklists in the text editor, tone matrix as a checklist before publishing, review of real examples rather than abstract principles, and including the same rules in prompts for AI-generated content, which increasingly writes drafts on behalf of people.
Investment in such a system pays off not only aesthetically: according to Frontify, brands with consistent communication earn 23% more than competitors with a misaligned voice — that's the argument to show to the CFO when he asks why another document about text tone is needed.
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
