Survey Design Kit

A survey does not measure your market — it measures the people willing to answer you, and everything honest you say afterwards begins by admitting that.

What it is

Most surveys fail before the first question is written, at the moment nobody said which decision the answer would change. This forges the whole instrument: the test that kills the survey which should never be fielded, a frame whose biases are directions not disclaimers, a questionnaire free of the four biases that manufacture the answer, and a report that says what a small base cannot carry. For marketers, founders, product leads and anyone about to email a list and call the result evidence.

What you receive

a survey that produces a decision instead of a chart — the decision it serves and the verdict on whether a survey is the right instrument at all, the frame with its biases named in the direction they push, a questionnaire clean of the four biases that ruin most surveys, the scale, length and order decisions, the fielding and reminder plan with its abort check, and a report stating base sizes, read bands and what the data cannot carry

Delivered as a single, complete .md skill file — open it, paste it into any capable chat model, and work. No installation, no account with anyone but us.

Supported models

• Claude (Opus/Sonnet)
• ChatGPT (GPT-4o/5)
• Gemini
• any capable chat model

People reach for it when

• “how to write survey questions that are not leading”
• “how many responses do I need for my survey to mean anything”
• “only 40 people answered my survey can I still use the results”
• “should my rating scale have a neutral middle option”
• “customer survey that actually tells me something I can act on”

Record

In the collection. The sale room is being fitted; this work goes on sale when it opens.

License

Free-shelf works ship under the DolceDrago House License — use the outputs freely, personal or commercial; the work file itself stays with the house.