The Translator Who Learned the Words But Not the Language.
Brands have more customer data than ever. They use it to personalise the words. Jenny Thomas identified the failure that happens next — in 1983, before MarTech existed.
In 1955, a Japanese electronics company called Tokyo Tsushin Kogyo was trying to rename itself for the American market.
The company's founders — Masaru Ibuka and Akio Morita — understood that their existing name was unpronounceable and unmemorable to English speakers. They wanted something short, global, and easy to say in any language. They consulted a Latin dictionary and found the word sonus, meaning sound. They combined it with the American slang word sonny — which in 1950s American English carried connotations of brightness, youthfulness, and energy.
The new name was Sony.
What Ibuka and Morita understood — and what most brands operating across contexts do not — is that language is not a system of words. It is a system of meanings. And meanings are embedded in context, in relationship, in the unstated assumptions that speakers share without needing to articulate them. A translator who knows every word in a dictionary but has never lived inside the culture that produced those words will construct sentences that are technically correct and experientially wrong.
The sentence is right. The moment it was sent into is wrong. And nothing in the words themselves reveals that.
Most brand personalisation is this translator.
The Research That Named This Failure
In 1983, a British linguist named Jenny Thomas published a paper in Applied Linguistics — Volume 4, Issue 2 — that gave precise language to something every failed cross-cultural business negotiation, every tourist who offended without knowing how, and every brand communication debrief has in common.
She called it pragmatic failure: the inability to understand what is meant by what is said.
Thomas's finding was that most communication breakdowns are not failures of vocabulary or grammar. They are failures of something more fundamental — the ability to use language appropriately in its social context. She distinguished two specific types.
The first is pragmalinguistic failure — using the wrong linguistic form for a particular communicative function. A request phrased as a command. An apology constructed as an explanation. The words are present. The social function they are performing does not match what the situation calls for.
The second — and the one that governs almost every brand communication failure — is sociopragmatic failure. This occurs not from a linguistic error but from a misjudgment of the social conditions for communication. The speaker has correctly identified the words. They have misjudged what the moment required. They are speaking into a relationship they have not correctly read.
Thomas was writing about human speakers navigating cultural boundaries. Her framework describes, with extraordinary precision, what happens every time a brand addresses a customer using data without understanding context.
The brand that sends a renewal reminder on the day a customer's parent died is not making a linguistic error. Every word is correct. Every product is relevant. It has committed a sociopragmatic failure — it has misjudged the social conditions for communication so completely that the correctly personalised message lands as an intrusion into a moment that required silence.
Unlike the cross-cultural traveller who eventually learns the unwritten rules of a new language through years of lived experience, the brand's data system has no mechanism for learning what the moment required. It only knows what the calendar said.
The Difference Between Speaking a Language and Living in It
Knowing the words of a language is necessary for communication. It is not sufficient.
A fluent non-native speaker can construct sentences that no native speaker would find technically incorrect. And yet something is always slightly off — a register that does not match the situation, a formality level that misreads the relationship, a reference that lands in the wrong emotional register. The gap is not in the vocabulary. It is in the accumulated, largely unspoken knowledge of what words mean in context — knowledge built through years of lived experience inside a language, not through studying its rules from the outside.
The American sociolinguist Dell Hymes named this gap in 1966. He called what native speakers have — and non-native speakers must develop over years — communicative competence: the ability to use language appropriately in social context. Knowing when to speak. Knowing what a particular situation calls for. Knowing how to say something in a way that respects the relationship between the people in the conversation. Knowing what words will mean to the person receiving them given everything that person already knows and is currently experiencing.
The most sophisticated customer data platforms available today can construct a message with a customer's name, location, recently viewed products, last purchase date, and the channel they responded to most recently. They can segment the customer into a behavioural cluster with impressive precision. They can predict, with measurable accuracy, which offer the customer is most likely to respond to at which time of day.
They cannot know what the moment requires.
Not because the data is insufficient. Because data describes behaviour without understanding the human conditions that produced it. The platform has learned the words of the customer's language. It has not lived inside it — and no amount of additional data closes the gap between those two things, because the gap is not a data problem. It is the gap between linguistic knowledge and communicative competence that Hymes spent his career describing.
The Four Ways Pragmatic Failure Shows Up in the Campaign Calendar
"Our Personalisation Is Working — Our Customers Say It Feels Irrelevant"
The first symptom of sociopragmatic failure at scale.
Most MarTech personalisation operates at the pragmalinguistic level — the right word, the right product, the right name, the right time of day. What it rarely reaches is the sociopragmatic level — the right relational register for what has just happened in this customer's relationship with this brand.
A customer who had a poor service experience last week receives an upsell email this week, addressed by first name, with a recommendation based on their purchase history. Every word is correct. Every product is relevant. The sociopragmatic conditions — what this moment requires given what just happened between this brand and this person — were never consulted.
The customer who says the personalisation feels irrelevant is not responding to the words. They are responding to the gap between what the words assume and what the moment actually is.
The translator who correctly conjugated every verb in a condolence letter that opened with "exciting news" has achieved pragmalinguistic success and sociopragmatic failure simultaneously.
"We Increased Personalisation Depth and Engagement Dropped"
The second symptom — the one that should force a reexamination of what personalisation is actually doing.
A brand invests in more granular segmentation. More data points per customer profile. More behavioural signals incorporated into the recommendation engine. The personalisation becomes more accurate, in the sense that it is more precisely tailored to documented behaviour.
Engagement drops.
The explanation offered in the debrief rotates through data quality, creative execution, and send timing. The explanation almost never offered is the correct one: the more precisely the system tailors a message to a documented behaviour pattern, the more clearly it communicates that this message was constructed by something that knows what the customer did and has no idea who they are.
Communicative competence requires knowing when to speak. A speaker who addresses every topic in the customer's file, with increasing accuracy, is not demonstrating communicative competence. They are demonstrating access to the file.
The tutor who covers every subject the student listed on their intake form has followed the rules of tutoring. The tutor who noticed what the student did not say has learned the language.
"We Localised the Campaign but the Market Didn't Respond"
The third symptom — the cross-cultural version of pragmatic failure that is most recognisable but least understood.
Localisation, as most brands practice it, is pragmalinguistic. The words are translated into the local language. The imagery is adapted. The product names are adjusted. What is rarely localised is the sociopragmatic framework — the unwritten rules about what communication between a brand and a customer means in that specific market.
In 2006, HSBC ran a global campaign under the tagline "Assume Nothing." The tagline was correctly translated into multiple languages. In several markets the translation rendered the phrase "Do Nothing" — a pragmalinguistic error with sociopragmatic consequences. The bank spent $10 million rebranding to "The World's Private Bank" in the markets affected.
The words were translated by professionals. The meaning was lost before anyone who lived inside those languages had been consulted. Thomas's sociopragmatic failure — the misjudgment of the social conditions for communication — is not prevented by translation competence. It requires communicative competence in the target context, which is built only through genuine presence in it.
"Our AI-Generated Content Tests Well but Feels Off to Our Best Customers"
The fourth symptom is the most contemporary and the most instructive.
AI-generated marketing content is extraordinarily capable of pragmalinguistic performance. It can match tone, vocabulary, sentence length, and stylistic register with impressive accuracy. A well-prompted language model produces content that passes every surface linguistic measure.
What it cannot replicate is communicative competence as Hymes defined it — the knowledge of when to speak, what this moment requires, what these words will mean to this specific person given everything they are carrying that the data does not contain.
The customer who says the AI content feels slightly off is not identifying a grammatical error. They are detecting a sociopragmatic failure — a message that has correctly learned the words of their language without ever having lived inside the conditions that produced them.
This is not solvable through more training data. It is a structural property of the gap between linguistic knowledge and communicative competence. The gap does not close through more accurate representation of past behaviour. It closes only through genuine engagement with the conditions of the present moment — which is exactly what a language model trained on historical data is least able to provide.
What Communicative Competence Actually Requires of a Brand
Hymes argued that communicative competence is not innate — it is built through sustained, attentive engagement with the social conditions of communication, with genuine feedback that updates understanding based on what each moment revealed.
Four practices that begin to build it at scale.
Practice 1 — Map the relational state before the content state
Before selecting content for any communication, the first question should not be "what is this customer most likely to respond to?" It should be "what is this customer's current relational state with this brand?"
A customer who just contacted service has a different relational state from one who has been quietly engaged for six months. A customer who just received a discount has a different relational state from one who paid full price yesterday. The relational state governs what the moment requires. The content is secondary.
Practice 2 — Build sociopragmatic suppression rules into the campaign system
Most suppression logic is behavioural — suppress if purchased in the last 14 days, suppress if complaint rate above threshold. Sociopragmatic suppression is different. It asks: what has just happened in this relationship that changes what this moment requires?
A service complaint filed in the last 72 hours is a sociopragmatic signal. A product return processed yesterday is a sociopragmatic signal. A fraud report raised this morning is a sociopragmatic signal.
None of these necessarily mean the customer should not be communicated with. They mean the communication that follows requires a different register — one that acknowledges what just happened before it asks for anything.
Practice 3 — Measure register, not just relevance
Relevance is a pragmalinguistic measure — did the content match the customer's documented preferences? Register is a sociopragmatic measure — did the tone, timing, and relational framing match what the customer's current situation required?
Build a monthly register audit: pull 50 customer communications at random, identify the relational context in which each was sent (recent complaint, recent purchase, long dormancy, first interaction), and evaluate whether the communication's register matched that context. No analytics tool does this automatically. A person reading 50 emails against their relational context in two hours will identify patterns that no dashboard surfaces. Do this once a month. The patterns that emerge are the sociopragmatic failures the system cannot see.
Practice 4 — Treat silence as a sociopragmatic choice
Communicative competence includes knowing when not to speak. Not because there is nothing to say — because the moment does not call for speaking.
Most campaign automation has no concept of choosing silence. It has suppression rules — passive conditions that prevent a send. Choosing silence is active — the communication could have happened, all conditions were met, and the assessment of the relational moment concluded that silence was the right choice.
The brand that can make this choice deliberately, on the basis of a read of the relational moment rather than a failed condition in a suppression rule, has begun to develop something that approaches communicative competence.
The Name That Felt Right Before It Needed to Be Explained
Sony succeeded not because Ibuka and Morita found a word that translated correctly into every market. They found a name that felt right before it needed to be explained — carrying its Latin clarity and its American warmth simultaneously, asking nothing of the listener except to say it.
This is communicative competence at the level of brand identity. The name that knew what it meant in context before any market had been asked to teach it.
The campaign that knows what the moment requires before the customer has had to say so is the same achievement at the level of individual communication.
Most brands are still learning the words.
The ones worth paying attention to are learning the language.
Not what to say. What this moment requires.
If your brain is already triaging this page for a 5-second window, skip the reading—the complete narrative is perfectly laid out in the infographic below.
Published by Hetvabhas — independent analysis of brand communication
infrastructure. No vendor agenda. No sponsored content. No false reasoning.
