aethrox To Questionnaire
A questionnaire only earns its place when the knowledge gap is real and the recipient is the only one who can close it; this skill's defining constraint is that every question in the document targets that gap specifically, grounded in established survey-design practice: neutral, unbiased phrasing (never a leading or loaded question), the right choice between closed and open questions for what's actually being asked, and a correctly-balanced scale wherever a rating is needed.
Interview the person sending the questionnaire only about the send (who it's going to and what's needed back), never about the subject itself, since the subject is exactly what the recipient knows and the sender doesn't.
Phase 1: Scope the send
Establish two things before drafting a single question:
- Who's the recipient: their role, expertise, and relationship to the sender. This fixes the questionnaire's tone and how much background context it needs to carry; a peer needs less framing than someone outside the team entirely.
- What's actually needed back: the specific decisions or facts the sender can't resolve alone. A vague goal ("get their thoughts") produces a vague questionnaire; a concrete list of what the sender must walk away able to decide produces a targeted one.
Phase 2: Choose closed or open per question, deliberately
Neither is a default; each question earns its format from what it's actually collecting:
- Closed (multiple choice, yes/no, a scale) when the answer needs to be quick, comparable across respondents, or clearly structured: at the cost of not capturing anything outside the offered options.
- Open when the sender needs context, reasoning, or an answer that can't be enumerated in advance: at the cost of being harder to compare or aggregate later.
Pairing a closed question with a short open follow-up ("why?") is often the right shape when both the fact and the reasoning behind it matter.
Phase 3: Write every question to avoid leading and loading
- No leading language. "How satisfied are you with X?" gets an honest answer; "Do you agree that X is working well?" pulls the respondent toward agreeing. Neutral phrasing produces the respondent's actual view, not the one implied by the question.
- No loaded or double-barreled questions. One idea per question: a question that bundles two things ("is the API fast and reliable?") produces an answer that doesn't actually resolve to either one.
- No compound assumptions. A question that presumes an unstated fact ("how much did the outage cost you?" presumes there was measurable cost) traps a respondent who'd otherwise say the premise is wrong.
Phase 4: Balance any rating scale
Where a question needs a rating rather than a free answer, build a proper Likert-style scale, not an ad hoc one:
- Offer equal numbers of positive and negative options around a genuinely neutral midpoint; "Excellent / Very Good / Good / Fair / Neutral" is unbalanced (four positive-leaning terms and no real negative); "Excellent / Good / Neutral / Bad / Poor" is balanced.
- Label the endpoints (and ideally every point) so respondents share a common understanding of what each value means, rather than each anchoring the scale differently in their head.
- Five points is a reasonable default, enough resolution to be useful, not so many that adjacent points become indistinguishable.
Phase 5: Assemble the document
Order questions most-important-first: an async questionnaire may only get one pass before the recipient stops. Group under theme headings once there are more than a handful. For each question: state it as one idea, leave an answer stub directly beneath, and add a one-line "why this matters" only where the question could otherwise be misread or invite a throwaway answer. Open with a short context paragraph orienting a recipient who wasn't in the sender's head, and close with a catch-all for anything not asked but worth knowing. Write the questionnaire without em dashes, using a comma, colon, period, or a reworded sentence instead.
Done when
- The recipient and the specific gap they're being asked to close are both stated before any question was drafted.
- Every question is closed or open on purpose, not by default.
- No question leads, loads, or bundles more than one idea.
- Any rating scale is balanced around a genuine neutral midpoint with labeled anchors.
- Questions are ordered most-important-first, one idea per question, with context up front and a catch-all at the end.