Tone of voice (tone of voice) as part of brand style
Voice, tone and style — why separate these concepts in practice
Confusion between these three words is the main reason why ToV guidelines don't work. Voice is the brand's personality, it does not change: if the brand is bold and witty, it remains so in advertising, as well as in a letter of apology for a failure. Tone is the adjustment of the voice to the situation: the same bold brand in crisis communication softens the humor, but does not completely abandon its intonation. Style is already specific tweaks: punctuation, sentence length, use of emojis, addressing by 'you' or 'you (formal)'.
The mistake of most teams is mixing all three levels into one document with adjectives like 'friendly, expert, lively'. As a result, copywriters do not understand what exactly they should do with these words in a cancellation letter. Separating into three levels provides operational clarity: voice is described once and almost never reviewed, tone — in a table for each channel, style — in specific text formatting rules.
Why restrictions work better than a list of beautiful adjectives
This is where the main gap between the 'showcase' and the working document lies. The showcase ToV-guide lists the qualities a brand should have: bold, warm, expert. The problem is that an adjective does not tell the author of the text what specifically must not be written. Two copywriters will interpret 'bold' differently, and the results will diverge so that the reader will notice the inconsistency.
Practice of large product companies confirms this. HubSpot and Grammarly do not ask teams to memorize a list of adjectives in their brandbooks — they require uploading a specific list of words that the brand never uses, and a sample text of 500 words written in the correct voice. Mailchimp went even more specific and explicitly stated in their guidelines: 'we do not call people ninjas, rock stars, or wizards' — a phrase that instantly eliminates an entire category of marketing clichés without lengthy explanations.
The Russian example of the same principle is the open editorial policy of Dodo Pizza. There is no abstract 'we are friendly,' but instead there is a strict dictionary rule: not 'users,' not 'customers,' not 'visitors' — only 'guests' or 'clients.' This rule is transferred to real texts verbatim, because it is impossible to interpret it ambiguously. It is this kind of specificity that distinguishes the document that the team actually opens before starting work, from the document that lives in the folder 'Brandbook_final_v3'.
One voice — different tones: how it works with Alice
Let us take the case of Yandex as a telling example of how the matrix platform theory works at scale. Alice is embedded in search, smart speakers, navigation, banking apps, delivery — dozens of contexts with completely different user tasks. The character's voice remains unchanged: light irony, lively conversational speech, absence of jargon. But the tone changes depending on the situation.
When a user asks for the weather forecast, the tone is playful and concise. When it comes to financial operations or privacy settings, irony takes a back seat, and the responses become more accurate and terse — because the cost of misinterpretation is higher. At the same time, recognizability is not lost: the user cannot confuse Alice with a faceless voice assistant of a competitor, even when she is serious. This is the essence of operational ToV — not 'to be funny all the time,' but to have a range within which the voice remains true to itself.
The 'platform × tone' matrix as a working tool
To replicate this effect without the scale of Yandex, it is sufficient to build a simple table for at least four channels — website, social networks, email newsletters, and support — and define for each a permissible range along the tone axes.
| Channel | Formality | Humor | Audacity | Enthusiasm |
|---|---|---|---|---|
| Website / landing | moderately formal | minimum | restrained | medium |
| Social Media | informal | actively | acceptable | high |
| Email newsletters | average | measuredly | restrained | medium |
| Support | formal in a crisis, otherwise — average | excluded | excluded | low, but not dry |
Axes are taken from the Nielsen Norman Group model: formal ↔ everyday, funny ↔ serious, respectful ↔ audacious, enthusiastic ↔ neutral. An important nuance is that these are independent scales, not one line of 'friendliness'. A brand can be simultaneously informal and completely non-audacious, or serious, but enthusiastically devoted to the topic. It is precisely the combination of four points, not a single word, that defines a unique position.
Errors in implementation that destroy trust
The most common and most expensive mistake is inconsistency between channels. The brand is friendly in social networks, formal on the website, and dry in support — as if these were three different companies. The audience does not formulate this as a 'ToV violation', they simply stop trusting: a feeling arises that behind the beautiful posts stands an indifferent support service, and vice versa.
- The guideline is written with adjectives without examples — each author interprets them in their own way.
- ToV was developed by marketers apart from copywriters, SMM specialists, and customer support — nobody opens the document.
- There is no stop-list of words — the team continues to use clichés that the brand declaratively rejects.
- The tone is not adapted for crisis situations — in a force majeure, the brand continues to joke, which reads as disrespect for the client's problem.
- The document is not updated — new channels (for example, messengers) appear, but there are no rules for them, and employees invent them on the fly.
There is a simple set of red flags that can be used to diagnose the problem without a deep audit: low engagement with externally 'correct' content, clients cannot describe in two words 'what kind of' character the brand has, different tones across channels cut the eye, and in the media space the brand is not recognized among competitors even without a logo.
The structure of the guideline that the team will actually use
The working document is built not 'top-down' from the mission to adjectives, but from the tool to the theory. Practice shows that the order of sections itself determines whether they will be used.
- The stop-list of words and expressions — the first section, because it is opened most often before writing text.
- A glossary of replacements: 'how we call this or that' — following the example of Dodo with 'guests' instead of 'users'.
- A matrix of 'platform × tone' with ranges across four axes NN Group.
- Samples of ready-made texts from 300–500 words for each channel — not abstract principles, but texts that can be copied structurally.
- A separate protocol for crisis communications — what can and cannot be written when the service crashes or a client is publicly dissatisfied.
- A short description of the voice on one page — for new employees and agencies; without it, the document looks disconnected from the brand's mission.
This approach unfolds the logic of a showcase brandbook: there theory comes first, and examples are in the appendix, if there are any at all. In the working document, examples and prohibitions are the core, while the theoretical description of the voice is a brief appendix for context. The team opens the document not to be inspired, but to quickly find the answer to the question 'can we write it this way or not' — and finds it in ten seconds, not after reading the section on brand values.
Voice, tone and style — why separate these concepts in practice
Confusion between these three words is the main reason why ToV guidelines don't work. Voice is the brand's personality, it does not change: if the brand is bold and witty, it remains so in advertising, as well as in a letter of apology for a failure. Tone is the adjustment of the voice to the situation: the same bold brand in crisis communication softens the humor, but does not completely abandon its intonation. Style is already specific tweaks: punctuation, sentence length, use of emojis, addressing by 'you' or 'you (formal)'.
The mistake of most teams is mixing all three levels into one document with adjectives like 'friendly, expert, lively'. As a result, copywriters do not understand what exactly they should do with these words in a cancellation letter. Separating into three levels provides operational clarity: voice is described once and almost never reviewed, tone — in a table for each channel, style — in specific text formatting rules.
Why restrictions work better than a list of beautiful adjectives
This is where the main gap between the 'showcase' and the working document lies. The showcase ToV-guide lists the qualities a brand should have: bold, warm, expert. The problem is that an adjective does not tell the author of the text what specifically must not be written. Two copywriters will interpret 'bold' differently, and the results will diverge so that the reader will notice the inconsistency.
Practice of large product companies confirms this. HubSpot and Grammarly do not ask teams to memorize a list of adjectives in their brandbooks — they require uploading a specific list of words that the brand never uses, and a sample text of 500 words written in the correct voice. Mailchimp went even more specific and explicitly stated in their guidelines: 'we do not call people ninjas, rock stars, or wizards' — a phrase that instantly eliminates an entire category of marketing clichés without lengthy explanations.
The Russian example of the same principle is the open editorial policy of Dodo Pizza. There is no abstract 'we are friendly,' but instead there is a strict dictionary rule: not 'users,' not 'customers,' not 'visitors' — only 'guests' or 'clients.' This rule is transferred to real texts verbatim, because it is impossible to interpret it ambiguously. It is this kind of specificity that distinguishes the document that the team actually opens before starting work, from the document that lives in the folder 'Brandbook_final_v3'.
One voice — different tones: how it works with Alice
Let us take the case of Yandex as a telling example of how the matrix platform theory works at scale. Alice is embedded in search, smart speakers, navigation, banking apps, delivery — dozens of contexts with completely different user tasks. The character's voice remains unchanged: light irony, lively conversational speech, absence of jargon. But the tone changes depending on the situation.
When a user asks for the weather forecast, the tone is playful and concise. When it comes to financial operations or privacy settings, irony takes a back seat, and the responses become more accurate and terse — because the cost of misinterpretation is higher. At the same time, recognizability is not lost: the user cannot confuse Alice with a faceless voice assistant of a competitor, even when she is serious. This is the essence of operational ToV — not 'to be funny all the time,' but to have a range within which the voice remains true to itself.
The 'platform × tone' matrix as a working tool
To replicate this effect without the scale of Yandex, it is sufficient to build a simple table for at least four channels — website, social networks, email newsletters, and support — and define for each a permissible range along the tone axes.
| Channel | Formality | Humor | Audacity | Enthusiasm |
|---|---|---|---|---|
| Website / landing | moderately formal | minimum | restrained | medium |
| Social Media | informal | actively | acceptable | high |
| Email newsletters | average | measuredly | restrained | medium |
| Support | formal in a crisis, otherwise — average | excluded | excluded | low, but not dry |
Axes are taken from the Nielsen Norman Group model: formal ↔ everyday, funny ↔ serious, respectful ↔ audacious, enthusiastic ↔ neutral. An important nuance is that these are independent scales, not one line of 'friendliness'. A brand can be simultaneously informal and completely non-audacious, or serious, but enthusiastically devoted to the topic. It is precisely the combination of four points, not a single word, that defines a unique position.
Errors in implementation that destroy trust
The most common and most expensive mistake is inconsistency between channels. The brand is friendly in social networks, formal on the website, and dry in support — as if these were three different companies. The audience does not formulate this as a 'ToV violation', they simply stop trusting: a feeling arises that behind the beautiful posts stands an indifferent support service, and vice versa.
- The guideline is written with adjectives without examples — each author interprets them in their own way.
- ToV was developed by marketers apart from copywriters, SMM specialists, and customer support — nobody opens the document.
- There is no stop-list of words — the team continues to use clichés that the brand declaratively rejects.
- The tone is not adapted for crisis situations — in a force majeure, the brand continues to joke, which reads as disrespect for the client's problem.
- The document is not updated — new channels (for example, messengers) appear, but there are no rules for them, and employees invent them on the fly.
There is a simple set of red flags that can be used to diagnose the problem without a deep audit: low engagement with externally 'correct' content, clients cannot describe in two words 'what kind of' character the brand has, different tones across channels cut the eye, and in the media space the brand is not recognized among competitors even without a logo.
The structure of the guideline that the team will actually use
The working document is built not 'top-down' from the mission to adjectives, but from the tool to the theory. Practice shows that the order of sections itself determines whether they will be used.
- The stop-list of words and expressions — the first section, because it is opened most often before writing text.
- A glossary of replacements: 'how we call this or that' — following the example of Dodo with 'guests' instead of 'users'.
- A matrix of 'platform × tone' with ranges across four axes NN Group.
- Samples of ready-made texts from 300–500 words for each channel — not abstract principles, but texts that can be copied structurally.
- A separate protocol for crisis communications — what can and cannot be written when the service crashes or a client is publicly dissatisfied.
- A short description of the voice on one page — for new employees and agencies; without it, the document looks disconnected from the brand's mission.
This approach unfolds the logic of a showcase brandbook: there theory comes first, and examples are in the appendix, if there are any at all. In the working document, examples and prohibitions are the core, while the theoretical description of the voice is a brief appendix for context. The team opens the document not to be inspired, but to quickly find the answer to the question 'can we write it this way or not' — and finds it in ten seconds, not after reading the section on brand values.
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
