Two Hundred Survey Comments and One Honest Set of Themes

    survey analysis
    qualitative research
    customer research
    Two Hundred Survey Comments and One Honest Set of Themes

    Two hundred survey comments and no way to read them

    You ran the survey. The scores are in a neat chart. Then there is the open-ended box, where 214 people typed what they actually think, and it sits in a spreadsheet column nobody wants to open. So you skim the first twenty, pick the three quotes that agree with what you already believed, and move on.

    That is how good evidence gets wasted. The fix is not reading harder. It is coding the answers properly, and the prompt Code Open-Ended Survey Answers Into Themes You Can Defend does that in one pass.

    A short worked example

    Say you ask a gym's members: "What is the one thing we should change?" You paste in these responses:

    • "Always a queue for the squat rack at 6pm."
    • "Showers are cold half the time."
    • "Wish there were more evening classes, the 6pm one is always full."
    • "Staff are lovely but the app keeps logging me out."

    The prompt returns a codebook with definitions (Peak-time crowding, Facilities, Class schedule, App reliability), a table tagging each response, and then groups the codes into themes. Here, "Peak-time crowding" and "Class schedule" merge into one headline: evening demand outstrips capacity. It reports how many responses back each theme and marks any with fewer than three as tentative, so a single grumble about the app does not get promoted to a strategy.

    ThemeResponsesConfidence
    Evening demand outstrips capacity2 of 4Tentative at this sample size
    Facilities upkeep1 of 4Tentative
    App reliability1 of 4Tentative

    Where the value is

    Three things make this more than a summary. First, every theme is tied to verbatim quotes, so you can show a colleague the evidence instead of asking them to trust you. Second, the contradictions section keeps minority views visible, which is often where the interesting problems hide. Third, the "what the responses cannot tell me" note stops you claiming more than the data supports. A job that takes an afternoon by hand takes a few minutes, and the result is easier to challenge in a meeting.

    How to use it

    1. Export the open-ended column and paste the responses, one per line. Remove names and anything personal first.
    2. Fill in the survey question, who answered, and the decision you are trying to make. That last field shapes which themes get surfaced.
    3. Run it, then spot-check five tagged responses against the codebook. If a code feels loose, ask for it to be split or merged and re-run.

    Treat the output as a first pass from a fast analyst. You still own the judgment, but you start from a defensible structure instead of a blank page. Browse more research prompts in the Prompt-King library.