Validation rules
Validation decides what counts as an acceptable answer. It’s checked as the respondent answers and again when they try to move on, and an unmet rule blocks the page with a message beside the question.
Validation is separate from logic: logic decides whether a question is asked, validation decides whether its answer will do.
The rules
Required — the question must be answered. Available on every type respondents answer.
Min length and Max length — character counts on a text question.
Min value and Max value — bounds on a numerical question.
Min selections and Max selections — how many options must or may be ticked on a choice or list question.
Format — the answer must match a pattern. Presets cover email address, phone number, UK postcode, web address, letters only, numbers only, and letters and numbers only. There’s also a custom pattern option if you need something the presets don’t cover.
Match question — the answer must equal another question’s answer. The confirm-your-email case, on text, numerical and choice questions.
Required, used well
Required is the rule people reach for first and regret most often. A forced answer isn’t a true one — a respondent who doesn’t want to say either invents something or abandons the survey, and both are worse than a blank.
Better in most cases:
- Add an explicit “prefer not to say” choice and make the question required. You get a real answer either way, and you can tell a decline from an omission.
- Reserve required for questions where a missing value makes the whole response useless, typically screeners and anything routing later logic.
Questions that fill themselves in — meta, timing and captcha — have nothing for a respondent to do, so required doesn’t apply to them.
Format rules in practice
Use the presets before the custom pattern. They’re maintained centrally, so an email preset stays correct as edge cases turn up, while a hand-written pattern quietly rejects real addresses for the life of the survey. Keep custom patterns for genuinely local formats — a study ID, an internal reference.
Whatever you use, say in the question’s description what shape you’re expecting. A rejection with no explanation is where respondents give up.