The Schwung Brand Scan: a diagnosis, not a survey

Most brand scans produce a report you already half knew. You fill in what your organisation does, what your values are, who your audience is, and a few weeks later you receive a document containing your own answers in a tidier format. That isn't a diagnosis. That's a mirror that only reflects the surface.

The Schwung Brand Scan works differently, and this piece explains how.

A questionnaire measures your assumption, a diagnosis tests it

The difference doesn't lie in the length of the questionnaire or the quality of the questions. It lies in the starting position.

A questionnaire begins with your answers. A diagnosis begins with hypotheses that stand apart from what you think. Before you've filled in a single letter, the scan has already read your website and formed suspicions about the gap between what you show and what you mean. The conversation that follows tests those suspicions, not your self-image.

That is the essence of the hypothetico-deductive model, the reasoning approach that has been considered the gold standard for clinical reasoning in medicine for decades. Four steps: gather signals, form a hypothesis, interpret new signals, and adjust or confirm the hypothesis. A doctor who only asks what the patient feels and writes it down is not a diagnostician. He is a secretary. The Schwung Brand Scan is built on the first model, not the second.

Five professional lenses, each forming a suspicion

Before the conversation begins, the scan holds your website up against five professional lenses: branding, employer market, creative, web and marketing. Each lens forms its own hypothesis about what is going on.

The branding lens looks at whether your essence is visible in your communications, or whether two different organisations are speaking. The employer-market lens tests whether your careers story matches what your employees would say at a birthday party. The creative lens examines whether your output stems from a single foundation or from five separate decisions. The web lens watches whether your architecture gives the visitor an answer or makes him search. The marketing lens asks whether your reach matches the people you say you want to reach.

These five suspicions are the starting point of the conversation, not the outcome. They determine which questions are asked, and how deeply.

The conversation moves down from symptom to cause

Here lies the methodological core. The scan uses laddering as its probing method: a technique described in 1988 by Reynolds and Gutman in the Journal of Advertising Research, and a standard in brand research ever since. The principle is simple. An initial answer says something about behaviour or communication. The answer to "why?" says something about the choice behind it. The answer to the next "why?" says something about the conviction driving that choice.

Only at that third level, sometimes the fourth, do you reach the cause. Not what an organisation does or says, but what it believes.

Suppose an educational foundation enters in the scan that its communication is "too dull". A questionnaire registers that and offers a recommendation about tone or imagery. The scan moves down. On further probing, the tension turns out not to lie in the tone, but in the absence of a shared story about why the schools belong together. Each school tells its own story. The foundation has no story. That isn't a communication problem. That's a brand-foundation problem, and a different route.

Five guardrails keep the conversation honest

A language model that probes without guardrails can go on endlessly, reason in circles, or reach a conclusion too soon. That isn't useful as a diagnostic instrument. That's why five deterministic rules safeguard the conversation.

The turn cap ensures the conversation doesn't overrun. The anti-loop recognises when the same ground is being covered for the third time and steers around it. The coverage floor checks whether all five professional lenses have been addressed sufficiently before the conversation wraps up. The no-backtrack rule prevents conclusions already discussed from being reopened without new information. The laddering floor ensures the conversation doesn't stop at the first answer, but probes until the conviction underneath becomes visible.

These are not stylistic choices. They are technical rules encoded in software, separate from the language model. They exist because an AI without structure tends to please rather than diagnose.

The engine puts its reading to you, you correct it

At the end of the conversation, the scan formulates a core tension and a follow-up route. But it also asks whether that is correct.

That is not a politeness. It is member checking, a principle Lincoln and Guba described in 1985 as the most crucial technique for establishing credibility in qualitative research. The researcher shares his interpretation with the participant, so that errors can be corrected and additional information can be added.

In the scan, it works like this: the engine puts its reading to you, and you can confirm, add to or correct it. That correction feeds into the final diagnosis. Not as an overruling of the analysis, but as an additional signal. A diagnosis that isn't recognised by the person receiving it is not a diagnosis. It's an assumption in a smart jacket.

The same tool, a different diagnosis per organisation

The scan is not built to produce one type of outcome. The outcome follows from the tension the conversation exposes, and that tension differs from one organisation to the next.

In one organisation, the tension sits between what the website promises candidates and what the employees themselves would say about working there. Two stories, one brand. That points to an employer-brand question. In another, something different surfaces: the essence is in place internally, but the translation outward is missing. That points to an activation question, not a brand-foundation project.

It isn't the sector that determines the outcome. Nor the size of the organisation. What determines it is the relationship between what an organisation is and what it shows, and where those two diverge most sharply.

Richer material requires trust, not just technology

Here is the counterargument that needs to be taken seriously. Conversational AI yields richer answers than static questionnaires. That has been confirmed in research several times. But the same literature notes that a proportion of respondents reject conversational formats out of concerns about privacy and trust.

Richer material is therefore not an automatic consequence of a better instrument. The conversation only works if it builds trust. In practice, that means: transparency about what happens to the answers, no endless probing without a visible purpose, and an outcome that the respondent recognises as his own situation, not as a generic analysis.

That is why the member-checking step at the end is not an option but part of the method. Trust isn't built through a polished interface. It is built by showing that the diagnosis really is about you.

A diagnosis starts before the conversation

The Schwung Brand Scan is not a clever survey. It is a diagnostic instrument that begins with hypotheses, moves down through laddering, and concludes with a reading that can be corrected.

The difference from a questionnaire is not technical. It is methodological. A questionnaire asks what you think. A diagnosis tests whether that's correct, and looks for the cause when it isn't.

Communication that doesn't land as intended rarely has a communication problem. Anyone who wants to know that doesn't need a questionnaire. They need a diagnosis.

Want to know more about how Schwung examines brands? Read about the approach at schwungreclame.nl/aanpak or take a look at the Schwung AI Brand Scan.