Add to Slack The most common reason a survey result disappoints is that the question was the wrong shape. The team lead wanted to know if people were happy, asked “Are you happy?” as a free-text question, got forty paragraphs of nuance, skimmed three, and called the survey inconclusive.
The right shape was a 5-point agree/disagree on “I have been happy at work this quarter,” with a single follow-up free-text question for people who wanted to explain. The dashboard would have shown the distribution on the first chart and the nuance in the open-text on the second. Five minutes of analysis instead of an afternoon.
Picking the right question type is mostly about matching the shape of the answer you actually need. Here is the short field guide.
The six types, what each one is for
Free text is for nuance. Use it when the most useful answer is the one you did not predict. Use it sparingly. One free-text question per survey is usually the right budget; two is the upper limit before response rate drops.
Single-select is for a forced choice between options. Use it when the team has to land on one answer and you have already decided what the options should be. The hard work happens at question-design time, not at analysis time.
Multi-select is for shopping lists. Use it when the answer is a set: which tools, which formats, which channels, which weeks. Add a sensible “None of the above” or “Other” option whenever the answer might genuinely be empty; people who feel forced into a list will pick the least-wrong option and you will learn nothing.
NPS (0 to 10) is for sentiment at scale. Use it on teams of at least fifteen people. Below that, the score is noise; one detractor moves the number by 7 points. It is also a famously blunt instrument; pair it with a free-text follow-up so you know what the score is about.
Yes/No is for clear gates. Use it when there are only two answers and the in-between is a distraction. “Did the goals you committed to this quarter ship?” is a yes/no; “How well did the goals ship?” is a 5-point agree/disagree.
Agree/Disagree (5-point) is for sentiment with direction. Use it when “strongly” matters and you need to compare distributions across statements. “I felt I had enough time to do my best work” plus “I felt I was working on the right things” give you a 2D picture in two questions; the same pair as free-text would give you two paragraphs and no comparison.
A worked example
A quarterly engagement pulse for a team of forty. Six questions, four minutes to answer, a result the team lead can read in a single sitting.
- Yes/No. “Did the goals you committed to this quarter ship?” A clean answer the team lead can act on, and a useful denominator for everything else.
- 5-point agree/disagree. “I felt I had enough time to do my best work this quarter.” Direction plus magnitude. Pair with the next question for the workload vs focus cross-tab.
- 5-point agree/disagree. “I felt I was working on the right things this quarter.” Different question, same shape, designed for comparison.
- NPS. “How likely are you to recommend joining this team to a friend?” The single most-cited engagement signal, and a fine fit at this team size.
- Multi-select. “Which of these slowed you down most? Pick up to three.” A focused list with a sensible cap: too many meetings, unclear priorities, broken tooling, dependencies on other teams, none of these.
- Free text. “What’s the one thing you’d change about how we work?” The catch-all. Half the answers will be redundant with the earlier questions; the other half will be the most useful thing in the survey.
That mix gives you a quantitative spine (questions 1 to 4), a focused diagnostic (question 5), and a place for the surprises (question 6). The whole survey takes about four minutes to answer.
Mixing types beats running six separate polls
The most common alternative to a multi-question survey is to run each question as its own poll in a Slack message. It looks lighter and is not. The response rate drops on each subsequent message; by the third poll, half the team has stopped engaging. A single survey gets answered once, cleanly, with the same set of respondents on every question. The cross-tabs you can do at analysis time are only honest when the respondents are the same set.
The other reason: a multi-question survey shows up as one ping. Six separate polls show up as six pings. Your team will notice the difference.
A question-design checklist
Before you publish, run each question past three checks:
- Can you predict the most likely answer? If yes, ask a sharper question.
- If the answer is what you expect, will you do anything differently? If no, drop it.
- Could the question be answered with a number instead of a paragraph? If yes, change the type.
A survey that survives all three checks usually surfaces something worth knowing.