# Actionable Questionnaires: Start with the Decision, Not the Questions Build actionable questionnaires: link each question to a decision, then turn responses into concrete next steps. - Canonical URL: https://www.harmate.com/en/blog/actionable-questionnaires-start-with-the-decision-not-the-questions - Author: Harmate Team - Published: 2026-08-01 - Updated: 2026-08-07T16:41:20.602458+00:00 - Language: en ## Content # Actionable Questionnaires: Start with the Decision, Not the Questions An actionable questionnaire can be completed thoroughly and, above all, lead to a clear decision. You collect dozens, sometimes hundreds, of responses. You export them. You calculate an average. You produce a chart. Then the same question comes back: **what are we going to do with all this?** This guide is aimed first at HR and training teams that need to choose a concrete adjustment based on an internal survey or poll. The same method can then support product or research teams working with open-ended responses. The problem does not always come from the analysis. It often starts before the first question. A useful questionnaire is not a long list of topics we would like to cover. It is a system designed to shed light on a specific decision. Until that decision is clear, questions pile up, responses scatter, and the result becomes difficult to use. The method can be summed up in one sentence: **start with the decision, then work backwards to the questions needed to make it.** You will see how to apply a simple framework, review responses without manufacturing certainty, and connect every signal to a testable action. This article is an entry point. It complements [Open-ended questions: getting actionable answers without bias](/en/blog/open-ended-questions-get-actionable-data-without-bias), [Framing bias: why your questionnaires already contain the answers](/en/blog/framing-bias-why-your-questionnaires-already-contain-the-answers), and [Analyzing 200 open responses without bias: a method for actionable decisions](/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions). The goal is not to replace them, but to show how to connect them in one journey, from decision to action plan. A questionnaire can be long, widely distributed, and still provide nothing that helps you choose between two actions. Without that link, reading time accumulates, trade-offs are postponed, and the decision becomes difficult to justify. The difference is not the volume of responses. It is the decision the questionnaire is designed to prepare. ## Start by writing the decision Before writing a question, write down the decision the results must inform. It should be expressible with an action verb: - **adapt** training content; - **prioritize** product improvements; - **compose** working groups; - **support** a new hire's transition; - **understand** a gap between two populations; - **decide** whether an experiment should continue. A useful formulation looks like this: > At the end of the analysis, we must choose the two improvements to test in this journey. This sentence imposes discipline. It prevents you from mixing satisfaction, company culture, training needs, feature ideas, and topics that nobody will know how to address afterwards in the same questionnaire. If you do not know what decision will be made, you can still collect information. But you are not designing an actionable questionnaire yet. ## The CECA framework: the four types of information a decision needs A decision does not rest on an isolated score. It needs several types of signals, each with a different function. The **CECA** framework — **Context, Experience, Constraints, Action** — helps check that the questionnaire covers these four angles without losing focus. ### 1. Context The same answer does not mean the same thing in every situation. Ask what the person was doing, when it happened, and what constraints were present. **"In what situation did you use this process for the first time?"** **"Which step did you complete on your own, and which one did you complete with someone's help?"** Context prevents a one-off impression from becoming a general judgement. ### 2. Lived experience People do not always describe their experience with the same words as the organization. An open-ended question can surface what was actually understood, used, or felt. **"What seemed clear, difficult, or unnecessary to you?"** This wording allows several realities to coexist. It does not assume that the experience was positive or negative. ### 3. Constraints and preferences A preference is not a constraint. A constraint is not proof that a solution is impossible either. **"What would prevent you from using this solution in your usual context?"** **"Of the options proposed, which would suit you best and why?"** These answers help distinguish a non-negotiable need from a habit or a matter of convenience. ### 4. The next action A response is easier to use when it indicates what could be tried next. **"What change should we test first?"** **"What help would you need to take this step?"** The respondent does not decide on behalf of the team. They provide a starting point from which to build the action. ## Write without repeating the same mistakes For writing, keep three tests in mind: define a situation, ask for a fact before a conclusion, and leave room for the neutral or unexpected. The examples of reformulation are detailed in [Open-ended questions: getting actionable answers without bias](/en/blog/open-ended-questions-get-actionable-data-without-bias), while [Framing bias: why your questionnaires already contain the answers](/en/blog/framing-bias-why-your-questionnaires-already-contain-the-answers) addresses the risk of embedding an assumption in the question. An open-ended question is therefore not a vague question: it describes a situation precisely enough to produce a comparable answer without writing that answer in advance. ### Concrete examples: before and after **Example 1:** - ❌ *"Did you understand the role?"* - ✓ *"Describe what you understood about your role in the first week."* The first version invites a yes or no answer and provides no insight into what was actually understood. The second captures the lived experience and reveals gaps in clarity. **Example 2:** - ❌ *"Was the onboarding helpful?"* - ✓ *"What part of onboarding helped you move forward, and what slowed you down?"* The first assumes a binary experience. The second allows for mixed responses and surfaces the specific elements that matter, not just a general sentiment. ## What research best practices confirm This principle is not just an editorial preference. [AAPOR's best practices](https://aapor.org/standards-and-ethics/best-practices/) recommend pretesting a questionnaire and point out that wording, context, and question order can influence responses. The [Pew Research Center](https://www.pewresearch.org/writing-survey-questions/) also notes that small differences in wording can change how a question is interpreted. This does not turn a questionnaire into a perfectly neutral instrument. It simply provides a working rule: test important formulations before distributing them, then preserve the context needed to review the responses. ## Illustrative case study: clarifying onboarding The following example is deliberately illustrative: it is neither a customer report nor evidence of performance. It shows how to connect responses, context, and a decision. The decision is: **choose the two priority adjustments for an onboarding journey over the next three months**. Reading a response is not just a matter of counting the most frequent words. You need to connect three elements: **what is said, in what context, and what it means for the decision**. Consider three responses about an onboarding journey: - **"The information was available, but I did not know what to read first."** - **"I understood the role, but not how to ask for help."** - **"The first week was clear. After that, I no longer knew whom to ask my questions."** They do not necessarily call for a new manual. They may point to a problem with orientation and hand-offs at different moments in the journey. Useful analysis should group responses that address the same topic, preserve important exceptions, and connect each signal to the decision it is meant to inform. For large volumes, [Analyzing 200 open responses without bias: a method for actionable decisions](/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions) explains the sorting and review method in detail. Categorization helps reveal patterns. It does not automatically turn human nuance into definitive truth. ## Turn a signal into an action The CECA framework helps design the questionnaire. When reviewing the results, apply the same logic in a decision brief that answers five questions: - **What signal did we observe?** - **Which responses or situations support it?** - **What decision can it inform?** - **What action can we test?** - **When will we check whether this action changed anything?** In our case study, the signal does not automatically call for a new manual. It may instead lead you to test a clearly identified point of contact and a follow-up meeting on day 10. You can then check where new arrivals look for help after a month. This format avoids two common mistakes: producing a report that stops at the finding, or launching an action without being able to explain what motivated it. ## What a tool can handle — and what it must not decide A form collects responses. A spreadsheet can sort them. An analysis tool can help group themes, compare populations, identify gaps, and produce a clearer summary. In this decision-driven workflow, the questionnaire is only a starting point—the key is connecting it to a decision brief, analyzed responses, and testable actions. A tool's role is to make that journey visible and connected rather than scattered across documents. > **Checkpoint.** [Harmate Decision Files](/en/product/decision-files) embed this journey into your questionnaire workflow: you record the decision upfront, attach the questionnaires that will inform it, gather and analyze responses, and mark the actions tested—all linked in one place. This keeps the connection visible rather than scattered across documents. But a tool does not know your business constraints, political trade-offs, or the consequences of a decision on its own. Groupings, categories, and summaries must remain interpretable, open to challenge, and supervised by the people responsible for the decision. The quality of the result always depends on the quality of the questionnaire, the context shared with respondents, and the way the team reviews the signals. ## The actionable questionnaire checklist Before sending it, check that: **Design phase** - the decision to inform is written in one sentence; - every question has an identifiable purpose; - the questions describe a specific situation or period; - respondents can express a positive, negative, neutral, or unexpected experience; - preferences are separated from constraints. **Distribution phase** - the number of questions fits the attention available; - the collection respects the stated purpose and the level of confidentiality required. **Analysis phase** - the results will be reviewed by someone able to put them back into context; - a response can be connected to an action or a follow-up question. If you do not know what decision a question is meant to inform, remove it from the questionnaire or clarify the decision first. ## Frequently asked questions ### Should multiple-choice questions be removed? No. They are useful for quickly collecting comparable information. They become insufficient when they replace every opportunity to explain a choice, flag an exceptional case, or surface an unexpected need. ### How many open-ended questions should we ask? There is no universal number. For a short first version, you can start with two or three targeted open-ended questions, then check whether each one brings a nuance that could change the decision. If an answer will be neither read nor used, it is better not to ask for it. ### How many responses do we need before deciding? Enough to see patterns. For a team of 20, 15 responses often suffice. For 500 people, 80–100 reveal saturation and diminishing returns. The exact number depends on how homogeneous your population is. If the group is diverse, you may need more responses to capture different perspectives. The goal is not statistical perfection; it is confidence that you have heard the main signals. ### What if we get conflicting signals? Preserve both. Document them in the decision brief with the context that explains the difference. A conflict often reveals where circumstance matters—for example, remote workers may face a different barrier than in-office staff. This distinction should guide follow-up actions and shape the decision itself. Ignoring conflict can lead you to build a solution that works for one population but fails for another. ### Why not use Google Forms or Typeform with a ChatGPT prompt? A generic form collects responses, and a generic assistant can occasionally help summarize them. The difference is the workflow: the questionnaire, context, analysis, trade-offs, and follow-up remain connected in an explicit business journey. This does not remove human review or make a decision automatic. ### If nobody has time to review the results, is it useful at all? No. An actionable questionnaire must specify who will review the responses, what decision will be made, and when the action will be checked. If nobody owns that responsibility, reduce the scope or postpone the collection rather than produce another report. ### Can AI make a decision from the responses? It can help organize, compare, and summarize responses. It should not be presented as an authority that replaces human judgement. A decision remains tied to its context, consequences, and the people who are responsible for it. ### Can sensitive responses be analyzed? Yes, but only with a clear purpose, proportionate collection, understandable information for respondents, and appropriate access controls. A sensitive question is not justified simply because an analysis would be technically possible. ## Conclusion: do not collect a response you will not know how to use A useful questionnaire does not begin with a question. It begins with a decision. That decision gives every question a role, every response a context, and every observed signal a next step. It lets you move from a report that describes to a conversation that helps people choose. Without that link, reading time accumulates and trade-offs remain difficult to explain. With it, every response has a place in a decision journey, even when the final outcome is to change nothing. Open-ended responses become useful when they are structured, compared, and discussed in the context of the decision. **Before adding a question to your next questionnaire, write down the decision it is meant to inform.** If you want to move from that decision to a questionnaire, then to a readable analysis of the responses, [prepare your next decision with Harmate](/en/product/decision-files#decision-files-proof). Validation and trade-offs remain with your team.