survey-design

Designs unbiased survey questionnaires with analysis plans, sampling honesty, and pilot testing.

1|Updated Jul 3, 2026
One-click install
npx skills add https://github.com/Nandansai08/skillz --skill survey-design-nandansai08
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: survey-design
Source: https://github.com/Nandansai08/skillz/tree/main/skills/research/survey-design
Command: npx skills add https://github.com/Nandansai08/skillz --skill survey-design-nandansai08

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Surveys shipped without rigorous design produce permanently flawed data — leading questions, double-barreled items, and biased samples cannot be patched after 600 responses arrive. This Skill enforces a disciplined workflow so every question maps to a planned analysis and results are reported with honest sampling context. ## Core Features & Use Cases - Analysis-First Instrument Design: Work backward from a pre-registered analysis plan so every question maps to a planned finding, keeping surveys to 5–12 questions. - Question Wording Audit: Detect and fix double-barrels, leading frames, unbounded recall windows, and intention-based questions that predict nothing. - Sampling & Reporting Honesty: State the sample frame, response rate, nonresponse skews, confidence intervals, and exact question wording in every report. - Use Case: Quantify an interview finding across your user base — e.g., measure what percentage of new accounts delegate onboarding setup, segmented by company size, with a piloted 8-question instrument. ## Quick Start Help me design a survey to measure how common setup delegation is among new accounts, segmented by company size.

Frequently Asked Questions about survey-design

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I design a survey that isn't biased?▼

Start from an analysis plan so every question maps to a planned finding, then audit wording for double-barrels, leading frames, and unbounded recall windows. Pilot with 5–10 respondents thinking aloud to catch interpretation mismatches before fielding.

How many questions should a survey have?▼

Target 5–12 questions. Every added question costs response rate and completion quality, so a 40-question omnibus buys worse data on all 40. Cut any question that has no line in the analysis plan.

Should I use surveys or user interviews for research?▼

Use interviews first to discover unknown unknowns, then surveys to quantify what interviews surfaced. Surveys cannot reveal what you didn't know to ask; they measure prevalence of already-identified themes across a population.

Why are 'would you pay $X' survey questions unreliable?▼

Stated intent predicts nothing because saying yes costs the respondent nothing. Ask behavioral questions about past actions instead, or use real willingness-to-pay methods like Van Westendorp or conjoint analysis.

How do I handle low survey response rates and nonresponse bias?▼

State the sample frame honestly in the report: who was invited, who responded, and how respondents differ from the population. Nonresponse bias is worst when responding correlates with the topic, so describe skews and treat prevalence figures as bounded estimates.