Interviews 101

1. 01

   What makes a good interview?

   Design questions so results can be analyzed, compared, and acted on. Each question should measure a single, clearly defined idea using neutral, behavior-based language and consistent response scales. Avoid vague, emotional, or double-barreled questions.

   
   

   Write every item with the end analysis in mind—if you can't explain how responses would be interpreted or compared, the question isn't ready. Open-ended questions are valuable for context, but structured questions create the signal.

2. 02

   Interview Question Building Guidelines

   If a question can't be operationalized, it can't be trusted.

   If it can't be compared, it can't be learned from.

3. 03

   Design for analysis at the moment of writing

   Before finalizing any question, ask:
   - What variable is this?
   - What does "high" vs. "low" mean?
   - What decision could this inform?
   - What would we compare this against?

4. 04

   Measure constructs, not feelings

   Experts design surveys around constructs (e.g., role clarity, psychological safety, manager support), not emotional venting.

   
   

   Bad: "How do you feel about leadership?" (Answer is open ended)

   
   

   Good: "My manager gives me the information I need to do my job effectively." (Answer is a scaled answer, or Yes or No)

5. 05

   One idea per question. Always.

   Bad (testing two attributes): "My manager communicates clearly and supports my growth."

   
   

   Good:
   - Question 1: "My manager communicates expectations clearly."
   - Question 2: "My manager supports my professional growth."

6. 06

   Experts avoid vague terms like "often," "fair," or "good" unless anchored.

   Good: "In the past 3 months, I received actionable feedback from my manager." (Answer is Yes or No)

7. 07

   Avoid leading, loaded, or moralized language

   Bad: "My manager respects me."

   
   

   Good: "My manager listens to my input before making decisions that affect my work."

8. 08

   Use open ended questions to 'color' the picture but separate from quantitative signals

   Use closed-ended items for signal

   
   

   Use open text to explain why the signal exists