Most forms don't fail because the product behind them is bad. They fail because somewhere between question three and question eleven, the person filling them out decides it isn't worth the effort. Good form design is mostly about removing reasons to quit.
These are the principles we've found matter most, whether you build your form by hand or generate it with AI.
Every question costs you respondents. Before adding a field, ask: what decision does this answer feed? If you can't name one, cut it. "Nice to know" is how a five-question form becomes a fifteen-question form that nobody finishes.
A useful exercise: write down the decision the form supports ("shortlist 10 applicants," "decide what to build next quarter") and work backwards from it.
Order questions by effort, not by topic. Names, emails, and multiple-choice questions go first because they build momentum. Long answers, file uploads, and sensitive questions (budget, salary, phone number) go last, when the respondent is already invested.
The same applies to commitment: once someone has answered eight questions, they're far more willing to attach a resume than they would have been on screen one.
For anything beyond a handful of questions, split the form into pages, each focused on one thing: "About you," "Your experience," "Availability." Short pages keep each screen simple, which matters most on mobile, where a single wall of twenty fields is a guaranteed drop-off.
Multi-page forms also give you a progress signal for free: people finish things they can see the end of.
Field choice is quiet UX. A few rules of thumb:
Conditional logic is the difference between a form and an interrogation. If someone says they don't want a follow-up call, don't show them the phone number field. If they picked "student," skip the questions about their company. Every irrelevant question a respondent has to read (even to skip it) erodes trust that the form respects their time.
Small touches compound:
Assume a large share of your respondents are on a phone with one thumb and thirty seconds. That means: bigger tap targets, one column, short pages, minimal typing (choices over text where possible), and no question that requires switching apps to answer.
Test your form on your own phone before sharing it. If you sigh at any point, respondents will quit at that point.
Tell people what they're signing up for on the first screen: roughly how long it takes and what it's for. "Six quick questions, about two minutes" is a promise, and forms that make it get more starts and fewer abandonments than forms that hide their length.
Then keep the promise. Don't advertise two minutes and ask for an essay on page three.
Respondents are increasingly wary of where their answers go. Collect the minimum, say why you're asking for anything sensitive, and use a form tool that takes privacy seriously. Formly, for example, is hosted in Europe, GDPR-compliant, and doesn't use cookies or track respondents. If your audience is privacy-conscious, saying this on the form is itself a completion-rate feature.
Finally, design the after: route submissions somewhere they'll actually be seen (a webhook into your CRM, a notification to the right person) and act on them visibly. Respondents who see their feedback acknowledged fill out your next form too. A perfectly designed form that feeds an ignored spreadsheet is still a failed form.
The fastest way to apply all of this: describe your form to Formly. The AI drafts a structure that follows these principles, and you refine it from there. Or start from a template built for your use case.
