You don't need real user data — just the inputs that look valid and aren't. Run these cases against the signup form before your first real user does, field by field, from the first keystroke to the welcome email.

Email

IDActionExpected
01 Type user+tag@example.com Accepted
02 Type USER@Example.com Treated as the same account as user@example.com
03 Type user@@example.com Rejected under the field, not a toast
04 Type a leading/trailing space around a valid address Trimmed automatically, not rejected
05 Send the same values straight to the API Same rejections as the UI gives

Password

Password rules in the UI should match the server exactly — a password the form accepts but the API rejects is the classic "Something went wrong" a user can't do anything about. Paste should work without exception; blocking it is still disturbingly common and it breaks every password manager outright, which defeats the actual security benefit a stronger password was supposed to provide.

A 100+ character password should be accepted or rejected with a message, never silently truncated — the cruelest version of this bug lets someone sign up with a long password, truncates it to the first 72 characters server-side, and then rejects the full password on every future login with no explanation. If the form shows a strength meter, it should reflect the server's real rules, not an approximation that passes the UI and fails the API, or the reverse.

Name, phone, date of birth

Name fields need to render O'Brien, Müller, and Иванов as text, never as markup — and need to accept them at all, rather than silently stripping the apostrophe or rejecting non-Latin scripts outright. A name field that only accepts [a-zA-Z ] isn't a minor oversight for a global product; it's a field that can't be used by a large share of real names on earth.

Phone should accept a +, spaces, and a leading zero, and reject letters with a message rather than silently stripping them into a different number. Birthdate needs 29 February, a future date, and the exact 18th-birthday boundary all handled correctly — an off-by-one error on the age cutoff either locks out someone who turned 18 that same day, or lets through someone who turns 18 tomorrow.

Locale-specific formats for all three — name order, phone formatting per country, date display — are covered in more depth in the multilingual forms checklist; this page focuses on the validation logic itself rather than the formatting around it.

Duplicate accounts

Double-click the submit button, or press Enter twice, and expect exactly one account — not two, and not a server error on the second request that leaves the first one in a broken half-created state. Submitting the same email twice should be rejected identically whether it's an exact match, a different case, or a +tag variant your product treats as the same person — a case-sensitive uniqueness check is a common, quiet way for two accounts to end up sharing one real inbox.

A reused verification link should show a clear message ("already verified" or "this link has expired"), not a blank page or a generic error, and the browser's back button after a successful signup shouldn't resubmit the form and attempt to create a second account from cached form data.

The terms checkbox should actually block submit until checked — not just show a red asterisk while the button underneath stays clickable. A social-signup email that's already registered with a password needs a deliberate resolution: link the accounts, or explain clearly that the email is taken and offer the normal login instead — silently creating a second, disconnected account under the same email is the quietest version of the duplicate-account bug above.

The welcome/verification email itself is part of the form's test surface, not a separate concern: does it arrive within a reasonable time, does the verification link actually work on first click, and does an unverified account correctly get limited access (or none) until it's confirmed? A signup flow that "works" in the UI but silently fails to send the email leaves a user stuck with no feedback at all.

Don't

  • ✕Test only inputs you know are valid
  • ✕Block paste in the password field
  • ✕Trust the UI's validation without hitting the API directly
  • ✕Show "already registered" without deciding if that’s a privacy leak
  • ✕Reject names that use non-Latin scripts or an apostrophe
  • ✕Treat the welcome email as outside the form’s own test surface

Questions and answers

Yes. A plus sign is valid in the local part of an address and many people use it to filter mail. Rejecting user+tag@example.com is a bug, and a very common one.
It is a trade-off. The explicit message is friendlier, but it lets anyone check whether a given email has an account. If privacy matters, show a neutral message and email the address owner instead.
Because the truncated version is what actually gets stored and checked — a user who typed a 120-character password, unaware the server kept only the first 72, will have the full password rejected on every future login with no indication of why. The fix is to either support the full length or reject anything over the limit with a clear message at signup time.
The same as any other name — accept it, store it as entered, and display it back unescaped as text. A validation pattern restricted to ASCII letters silently excludes a large share of real names and should be treated as a bug, not a locale-specific edge case.
The account should still exist in a clearly "unverified" state with a visible way to resend the email, rather than leaving the user stuck with no account and no explanation. Test the resend path specifically, including what happens if it’s clicked several times in a row.
It's a reasonable anti-abuse measure, but it has to be generous enough to not punish shared IPs — offices, universities, and carrier-grade NAT on mobile networks can put hundreds of real, unrelated signups behind one address. Pair it with other signals (email domain patterns, velocity over longer windows) rather than a hard per-IP cap alone.
For ordinary fields — name, email, phone, birthdate, selects and checkboxes — yes, in one click. Passwords and captchas are better handled by hand.