Open-ended questions are powerful. They provide access to context, motivations, obstacles, and nuances. But when poorly asked, they produce vague answers that are impossible to compare and therefore impossible to exploit.
There is a common trap when wanting to "help" respondents: giving examples within the question. It is often counterproductive. The example biases the answer, reduces diversity, and injects your own vocabulary into the data. If the goal is to honestly analyze what people are saying, you must avoid suggesting the content.
This article gives you a simple method to write open-ended questions that generate comparable, useful, and directly actionable answers to form relevant groups or improve a training path, an onboarding, or an HR initiative.
A question fails when it is:
Too broad
Too abstract
Impossible to compare from one person to another
Biased towards "self-image"
Formulated without thinking about exploitation
The problem is not the openness. The problem is the lack of structure.
If a question measures several things at once, everyone answers in their own way, in their own order. You lose comparability.
You want to guide the quality of the answer, not its substance.
Do: Use constraints of recency ("recently", "the last time"), structure ("context, action, result"), or length ("3 to 6 sentences").
Don't: Give examples of situations or formulate a "right answer."
Opinions alone are hard to exploit. Facts are comparable.
Questions like "What are your strengths?" invite overselling. Instead, ask behavioral questions:
How the person organizes themselves.
How they react to a blockage.
What concretely wastes their time.
Before publishing, decide what you want to optimize:
Homogeneous groups: To facilitate learning pace.
Complementary groups: To cover different strengths.
You can paste this template under certain questions to structure without suggesting.
Context:
What I did:
Result:
What would have helped me:
Goal: Understand objectives, obstacles, pace, conditions for success.
Concrete Objective: "In a few weeks, what do you want to be able to do concretely thanks to this training? Describe the expected result in 3 to 6 sentences."
Context of Use: "In what context will you use what you learn? Describe the situation, framework, and constraints, remaining factual."
Main Blockage: "What is your main blockage today on this subject? Describe exactly what blocks you, without generalities."
Last Occurrence: "When did this blockage manifest for the last time? Describe the situation with: context, action, result."
Strategy When Blocked: "When you are blocked, what is your first reaction? Describe the sequence of what you do, in 4 to 8 lines."
Pace and Constraints: "What realistic pace can you maintain, and what constraints might slow you down? List 3 points maximum."
Goal: Collaboration, expectations, irritants, communication.
Functional Collaboration: "What makes a collaboration go well for you? Answer in 5 points max."
Dysfunctional Collaboration: "What signals announce that a collaboration is going to go wrong? List 3 to 5 signals."
Important Communication: "When a subject is important, how do you prefer information to circulate (channel, frequency)?"
Useful Feedback: "What is useful feedback for you? Give the characteristics: form, timing, precision."
Irritants: "What wastes your time the most in a team today? Describe the problem and its concrete consequences."
Performance Conditions: "In what conditions do you work best (framework, autonomy, ambiguity)?"
Look for similar signals: close objectives, compatible pace constraints, similar blockages.
Look for different strengths: different coping strategies, compatible communication preferences, "Architect" vs. "Builder" profiles.
A good open-ended question is not a "free" question. It is a question that leaves the content free but structures the response to make it actionable. By avoiding examples and imposing a neutral structure, you get authentic data.
Once collected, this high-quality data becomes the fuel for Harmate's analysis, allowing you to build coherent teams while keeping control over the criteria you want to optimize.