Read First, Then Ask

A survey can only find what its questions allow. The brief defines the question, the question defines the instrument, and the instrument defines what can be found. If the constructs in the questionnaire are not the ones people use, the data will be clean and the picture will be wrong.

The usual safeguard is a qualitative phase before the survey. It helps, but it inherits the same blind spot. Someone wrote the discussion guide from the brief, recruited against the team's hypotheses and moderated the room. The participants knew they were in a study.

Our answer is to read before anyone is asked anything. This is how that works.

1. Find where the category lives

Before a question is written, we identify the platforms and communities where people already discuss the category: Reddit threads, patient forums, comment sections, professional communities. These are places where no brand is present and no one is performing for a researcher.

2. Read to saturation

We collect unprompted conversation until new sources stop producing new patterns. Human analysts read it. The Reading Machine holds the whole conversation at once and shows where the volume sits, so the reading covers the edges as well as the centre.

3. Map the architecture

The output is a map of five things:

• the constructs people use to organise the category
• the tensions they carry
• the drivers behind their decisions
• the words they reach for
• the matters the conversation has stopped debating

4. Compare it with the brief

We set that architecture against the original brief and its assumptions. The gap between the two is usually where the most important findings already are. An attribute the team cares about may be absent from the conversation. A distinction consumers make constantly may be missing from the brief.

5. Write the instrument from the map

The questionnaire is then built on the architecture. Constructs reflect how people organise the category. Question wording mirrors their language. Variables that discovery surfaced go in, and variables that turned out to be absent come out. Answer options cover the positions people hold, which avoids the problem of a survey with no option that speaks for you.

6. Field it, and read the numbers against the map

The survey is fielded with the appropriate audience, and the results are interpreted against the qualitative reality that produced the instrument.

You can commission the whole study or only the discovery phase, as an input for your own team's quant work. Both are described in our method.