Most marketers assume the problem with data is a shortage. The real issue is more uncomfortable: every dataset is filtered through human intent before it reaches a dashboard. A recent ETBrandEquity piece traces that lesson through wartime intelligence, an unlikely car launch and a mathematical puzzle that refuses to be solved.
The interpretation gap
In 1941, during Germany’s invasion of the Soviet Union, commanders had reams of battlefield reports, yet the decisive question was how leaders read them. The article notes that Shakespeare captured the timing problem long before spreadsheets existed: “There is a tide in the affairs of men / Which, taken at the flood, leads on to fortune.” Military thinker Carl von Clausewitz later framed the same idea around the critical climax of an attack.
Fast forward to 2021. Western intelligence agencies, including the CIA and MI6, reportedly warned Volodymyr Zelenskyy that an attack on Ukraine looked imminent. But past false warnings, such as the pre-Iraq invasion claims, made the leader cautious. The data was present; trust and interpretation were not. Marketers live in a smaller version of this gap when dashboards show one story and stakeholders see another.
Expert reviews vs. real intent
The Jaecoo 7, a Chinese SUV, illustrates how data can contradict expert opinion. British motoring journalists were lukewarm about the vehicle despite its attractive price. Yet driver videos on TikTok generated strong consumer enthusiasm, and earlier this year the model became one of Britain’s best-selling cars. The behavioural signal from social video outran the editorial consensus.
That does not mean reviews are wrong or TikTok is truth. It means marketing teams need multiple lenses: stated preferences, revealed behaviour and context. If you only monitor the sources you have always trusted, you may miss the influx of demand building somewhere else.
Why “enough data” is the wrong question
In 1967, mathematician Benoît Mandelbrot asked how long the coast of Britain is. The Ordnance Survey figure is 17,820 km, but if you measure every pebble and inlet, the number grows without limit. His work on fractals showed that some questions do not have a final quantitative answer.
The parallel for marketing is direct: asking “how much data is enough” often ignores the more important question of what you are trying to decide. Philosopher Karl Popper observed that when you collect a certain type of data, you have already chosen a direction. Carissa Véliz adds a useful metaphor: what a mirror shows depends on how it is positioned. She argues that data is inert until narratives and relationships make it meaningful.
AI inherits the same biases
Training data comes from the web, a human artefact full of beliefs and blind spots. Researchers from Princeton University and the University of Chicago concluded that large language models “really are eager to create generalisations from limited data.” Even late last year, Yann LeCun, then Meta’s Chief AI Scientist, described a fork in artificial intelligence: one path relies on large language models, data centres and brute computing power, while the other looks toward world models and cause-and-effect learning. That choice cannot be resolved by data alone.
What marketers can do today
- Start with the decision and intent, not the dataset. Write down the question before opening the dashboard.
- Use at least three data lenses: expert or editorial reviews, declared consumer feedback, and revealed behaviour such as search, social mentions and purchase intent.
- Audit AI-generated insights by asking where the training data came from and which voices are missing.
- Treat “a single version of truth” as a decision process, not a final number. Regularly ask whose interpretation is shaping the narrative.
- Add small qualitative checks to big quantitative reports; one odd TikTok comment may reveal a shift before it shows up in market share.
Data-driven marketing is not the same as blindly following dashboards. The value comes from framing the right question, acknowledging bias and combining signals with judgement. That may sound less precise than a machine, but it is the only way to avoid building confident strategies on fragile interpretations.
Source: ETBrandEquity.com
