You sent the survey. Forty people responded. And now you're staring at a spreadsheet of numbers that tell you almost nothing. More often than not, the problem is not the topic but the survey question types you chose. The right format can turn the same topic into sharp, actionable data. The wrong one buries it. This guide covers the eight most-used survey question types with examples and tips on when each one earns its place.
For each type, you'll find when to use it, what it does well, and what to watch out for, covering training evaluation, employee feedback, event surveys, and customer research. The types range from closed-ended formats, where every answer option is predefined, to open-ended questions that capture free-text responses, and apply to everything from a quick pulse check to a full-length questionnaire.
1. Multiple Choice
Multiple choice questions provide a predefined set of answer options. Single-select versions force one choice; multi-select versions let respondents pick several.
Best for demographics, behavioral segmentation, and preference questions where the options are well-known. Examples:
- "Which training format do you prefer? (In-person / Virtual / Self-paced / Blended)"
- "Which of the following topics would you like covered in future sessions? (Select all that apply)"
Watch out for incomplete options that force respondents into an inaccurate answer. Always include "Other (please specify)" when you can't be certain you've covered every possibility. Randomize option order to reduce position bias, where respondents consistently pick the first or last answer.
2. Rating scales
Rating scales ask respondents to evaluate something along a numeric continuum, typically 1–5, 1–7, or 1–10. They quantify satisfaction, quality, and likelihood in a way that's easy to compare and track over time.
Best for measuring satisfaction, performance, and experience. The most common professional application is the Net Promoter Score (NPS), introduced by Fred Reichheld in a 2003 Harvard Business Review article [1]. NPS uses a 0–10 scale to measure likelihood to recommend. Examples:
- "How satisfied were you with today's workshop? (1–5)"
- "How likely are you to recommend this training to a colleague? (0–10)"
Always anchor your endpoints with clear labels: "1 = Very dissatisfied, 5 = Very satisfied." Unlabeled scales lead respondents to interpret numbers differently. For NPS, the standard scoring groups respondents as promoters (9–10), passives (7–8), and detractors (0–6) [1].
3. Likert scales
Likert scales present declarative statements that respondents rate on an agreement continuum, from "Strongly disagree" to "Strongly agree." They're technically a type of rating scale but common enough in professional surveys to treat separately.
Best for measuring attitudes, opinions, and perceptions across multiple related dimensions. They're well-suited to employee engagement surveys and training evaluations where you want to measure how strongly people feel about specific aspects of an experience. Example:
"The training content was directly relevant to my role."
Watch out for survey fatigue from too many Likert items in a row. Research on survey satisficing recommends keeping Likert grids to no more than 10 rows per block; beyond that, respondents tend to start selecting the same response for every item without reading carefully [2]. Break up long Likert sections with other question types.
4. Offene Fragen
Open-ended questions invite free-text responses without predefined options. They capture the qualitative context that structured questions miss.
Best for understanding the "why" behind quantitative data, surfacing unexpected themes, and capturing suggestions in respondents' own words. Examples:
- "What was the most valuable part of this session?"
- "What one change would most improve your experience?"
Limit yourself to one or two per survey and place them at the end, after structured questions. Completion rates drop on open-ended items because they require more effort. Analysis also requires thematic coding, which is time-intensive for large samples. For a quick workaround in live sessions, read responses aloud and ask the group to discuss the themes that come up most often.
5. Ranking questions
Ranking questions ask respondents to order items by priority, preference, or importance. Unlike multiple choice, they force trade-offs, revealing what matters most relative to other options.
Ranking questions are the forced-choice format of survey design. They stop respondents from quietly marking everything as "very important" and make them commit to an order, which is where the real priorities emerge.
Best for prioritization exercises and understanding relative preferences. Example:
"Rank the following training topics in order of importance to your role: AI literacy, communication skills, leadership development, cybersecurity awareness, data literacy."
Watch out for cognitive load. Ranking more than five to seven items becomes frustrating and produces unreliable data. Keep the list short and use clearly distinct options.
6. Matrix questions
Matrix questions use a grid format where rows are items and columns share a consistent scale. They let respondents evaluate multiple related items on the same scale without repeating the question format.
Best for comparing evaluations across multiple dimensions of the same experience. A practical example: a post-training evaluation matrix with rows for "Content relevance," "Trainer expertise," "Pace of delivery," and "Materials quality," each rated on a 5-point satisfaction scale. Four ratings, one screen.
Watch out for visual complexity on mobile devices. Large matrices are hard to read on small screens, leading to random responses. Keep matrices to five or six rows maximum and test on mobile before deploying to a wider audience.
7. Dichotomous (yes/no) questions
Dichotomous questions offer exactly two options, typically Yes/No or True/False. They're the fastest question format for both writing and answering.
Best for screening, filtering, and quick fact-checking. Examples:
- "Did you attend the full training session? (Yes / No)"
- "Have you used the new software since the training? (Yes / No)"
They work well as logic gates that route respondents to different follow-up questions based on their answer. A "No" to attendance could skip respondents past the session evaluation and send them straight to the demographic section.
Watch out for oversimplification. A "No" on its own tells you nothing about why. Follow dichotomous questions with a brief open-ended question or a rating scale to capture depth where it matters.
8. Demographic questions
Demographic questions collect background information about respondents, enabling segmented analysis. Common dimensions include department, role, tenure, location, and experience level.
Best for cross-tabulation and group comparison. Knowing that employees in one department rate training effectiveness significantly lower than the company average is far more actionable than a single overall score.
Make demographic questions optional, explain why you're collecting the data, and protect anonymity by avoiding segments small enough to identify individuals. A common threshold is five or more respondents per segment before reporting group-level data.
Putting it all together: designing a survey flow
A well-structured survey follows a logical progression. Start with simple screening or demographic questions to ease respondents in. Move to your core measurement questions: rating scales, Likert items, matrix questions. Close with one or two open-ended questions for qualitative depth.
Keep total completion time to five to seven minutes. Research from SurveyMonkey shows that surveys under seven minutes see significantly better completion rates than longer ones, with dropout rates rising sharply after 12 minutes [3]. That translates to roughly 10–15 questions depending on question type.
Häufige Fehler zu vermeiden
Even when you pick the right question types, a few design habits can quietly undermine response quality.
1. Asking double-barreled questions
A double-barreled question packs two separate ideas into one item: "The training content was relevant and the materials were easy to follow." A respondent who found the content relevant but the materials confusing has no honest answer. Split every compound statement into its own question. This is especially common in Likert-format surveys where writers try to be efficient by combining ideas.
2. Using leading language
Questions that point toward an expected answer inflate positive scores and make benchmarking useless over time. "How much did you enjoy today's session?" assumes enjoyment. "How would you rate today's session overall?" does not. Review every question for words that imply a preferred answer and replace them with neutral phrasing.
3. Offering unbalanced response options
A rating scale with four positive options and one negative one ("Excellent / Very good / Good / Average / Poor") is not balanced. Respondents who feel negative have limited choices, and your averages drift upward regardless of actual experience. For any scale that runs from negative to positive, match the number of options on each side of the midpoint. A standard 5-point scale works well: two negative, one neutral, two positive. If you don't want a neutral midpoint, use an even-numbered scale (4 or 6 points) that forces a directional response.
4. Burying the most important questions
Survey fatigue is real, and engagement drops as length increases. If your most important question sits at question 14 out of 15, a significant share of respondents may never reach it with full attention, or at all. Put your highest-priority questions in the first half of the survey. Save demographics and optional open-ended items for the end, where drop-off matters less.
Häufig gestellte Fragen
Wie viele Fragen sollte eine Umfrage enthalten?
For training evaluations and post-event feedback, 8 to 12 questions is a reliable range. That typically translates to five to seven minutes of completion time, which is where completion rates hold up best. If your survey is longer, consider whether every question will actually influence a decision. Questions that produce data no one will act on are worth cutting. A useful test before finalizing a survey: for each question, write down which decision or action it would inform. If you can't answer that, the question probably does not need to be there. This forces clarity about the survey's purpose before you send it out.
What is the difference between a rating scale and a Likert scale?
A rating scale assigns a number to an evaluation: "Rate this on a scale of 1 to 5." A Likert scale presents a statement and asks how much the respondent agrees with it: "The facilitator explained concepts clearly. (Strongly disagree to Strongly agree)." Both are ordinal scales, but Likert items are always paired with a declarative statement, while rating scales can apply to almost any evaluation task.
When should you use open-ended questions instead of structured ones?
Use open-ended questions when you don't yet know what categories matter, or when you suspect the structured options won't capture the full picture. They're valuable after a new program launch when you want to hear what stands out before you have enough data to build a rating framework around it. For recurring surveys on a stable topic, structured questions are faster to analyze and easier to track over time. A practical approach is to run one or two open-ended questions in the first round of a new survey, identify the themes that come up most, then convert those themes into structured options for future rounds.
Umfragen mit AhaSlides durchführen
Choosing the right question type only gets you halfway there. You also need respondents to actually engage and complete the survey.
AhaSlides is an all-in-one audience engagement platform that supports multiple choice, rating scales, open-ended responses, ranking questions, and more, all in a single live session. Build a 3-question mid-session pulse, display results on screen while the room is still together, and adjust your second half based on what you see. No separate survey tool required.
For training evaluations in particular, seeing the group's responses in the room changes the dynamic. The discussion that follows the data is often more valuable than the data itself. We've found that teams who see results together act on them, while teams who receive a summary report rarely do.
Quellen
[1] Reichheld, F. (Dezember 2003). „Die eine Zahl, die Sie zum Wachsen brauchen.“ (Harvard Business Review). https://hbr.org/2003/12/the-one-number-you-need-to-grow
[2] Krosnick, J. A. (1991). "Response strategies for coping with the cognitive demands of attitude measures in surveys." Angewandte Kognitive Psychologie, 5 (3), 213-236. https://doi.org/10.1002/acp.2350050305
[3] SurveyMonkey. "How long should a survey be?" https://www.surveymonkey.com/curiosity/survey_completion_times/
