---
name: survey-interview-design
description: Design or improve research surveys and interview guides with attention to measurement, participant experience, sampling, and pilot testing.
---

# Survey and interview design

Design an instrument that produces information suited to the question and can be used responsibly with the intended participants. Distinguish a draft instrument from a validated measure or an approved study.

## Define the measurement and setting

Identify the decision or research question, population, recruitment method, mode, languages, anticipated accessibility needs, duration, and planned analysis. Map each important concept to observable questions or prompts. Specify the relevant time frame and level of analysis; separate a participant's reported perception from a directly measured event. Ask for essential missing context or make labeled assumptions when a usable draft can still be prepared.

Choose an existing measure only when its construct, population, permissions, language, and administration conditions are suitable. Check the source and any modification rules. Changing wording, translation, response options, or mode can affect measurement; do not represent an adapted instrument as validated for the new use without evidence.

## Build the instrument

For surveys, write one clear idea per item, using familiar language and a defined reference period. Avoid presupposing experiences, double questions, leading formulations, and response options that overlap or omit plausible answers. Use skip logic only where it reflects the study's needs; define routing, nonresponse, and options such as not applicable when appropriate. Check that a scale's labels and direction are consistent, and avoid implying equal numerical intervals without justification. Match sensitive questions to a specific need and allow participants to decline when appropriate.

For interviews, select a structured, semi-structured, or open guide consistent with the research design. Prepare neutral opening questions followed by optional probes for examples, context, and divergent experiences. Keep probes distinct from questions whose wording must remain fixed. Identify how an interviewer will handle distress, requests to stop, and consent to recording; do not promise confidentiality or anonymity beyond the approved procedures.

Make the order workable and avoid premature disclosure of hypotheses when this would bias answers. If the same question is asked in different languages or modes, plan competent translation and cognitive checks with intended participants, rather than assuming literal translation preserves meaning.

## Test and deliver

Review each item against its purpose and planned analysis. Walk through every route, including missing and inapplicable responses, and estimate burden. Plan cognitive interviews or a small pilot with suitable participants to identify interpretation and accessibility problems; report pilot findings only if observed. Predefine how feedback will change the instrument and, for quantitative scales, plan appropriate reliability and validity assessment for the intended context. A small pilot is not by itself proof of representativeness, reliability, or validity.

Deliver the editable questionnaire or guide, item-to-objective mapping, administration and routing notes, analysis implications, unresolved assumptions, and a pilot plan. Identify recruitment, ethics review, consent, and data protections still requiring approval or local expertise. Use synthetic examples where real responses are unnecessary. Do not recruit participants, collect data, or transmit identifiable responses without the relevant authorization and applicable safeguards.

When revising supplied questions, give an editable item-level change record: original wording, issue identified, proposed wording or response options, and the reason for the change. Flag decisions requiring the researcher's input, and distinguish a proposed revision from a change tested with participants. Retain the original item when a meaningful revision cannot be justified without missing context.