Survey checker

Paste one question per line, or upload the file your survey lives in. Lines that are not questions (instructions, a thank-you) are kept but not scored. The example loaded by default shows the most common problems on purpose.

Read the findings from the top: high-severity items change what the data means; low-severity items change how many people finish.

• Length: more than eight questions in a transactional survey, or any question over twenty words.
• Wording: leading or loaded words (“our excellent service”), absolute words (“always”, “never”), internal jargon and abbreviations.
• Structure: double-barrelled questions (“friendly and fast”), duplicated questions, three or more scale types mixed, no closing line.
• Metrics: no core question (a 1–5 satisfaction or a 0–10 recommend question) and no open “why” question.
• Privacy: requests for phone numbers, ID numbers, date of birth, religion, health or income without a stated need.

Each rule carries a fixed penalty (high 12, medium 7, low 3 points), so the same survey always gets the same score. The rules are deliberately strict on wording: a survey that reads as neutral to the team that wrote it often does not read that way to a customer.

Apply the rewrites, cut to one metric question, two or three driver questions and one open question, then test it with twenty real customers before launch. If the comments repeat the question back to you (“what do you mean by…”), the wording is still wrong.

A good survey is only the start: the same questions need to run on the kiosk, the QR code, the SMS and the web link so the answers are comparable, and a low score needs to reach someone who can act. That is what Qmeter does; the survey template library holds ready-made versions of the surveys the checker approves of, and the guide to survey questions explains the reasoning behind each rule.