Codes and passwords are the bane of user experiences: even small mistakes can delay desired activities, causing undue frustration. Work on codes has focused on security instead of people’s ability to enter them error-free. Difficulties observed in a security demonstration motivated this investigation of code transcription difficulty. A pilot study with 33 subjects and a follow-up study with 267 subjects from 24 countries measured performance and preference for codes of varying lengths, patterns, and character sets. We found that, for users of all languages, long codes with alternating consonant - vowel patterns were more accurately transcribed and are preferred over shorter numeric or alphabetic codes. Mixed-case and alphanumeric character sets both increased transcription errors. The proposed CVC6 code design composed of six Consonant-Vowel-Consonant trigrams is faster to enter, more secure, preferred by users, and more impervious to user error when compared to codes typically used for security purposes. An extension integrates error detection and correction, essentially eliminating typos.