A transposed digit, a misheard postal code, a name spelled letter by letter: alphanumeric content is where production ASR still fails in ways that cost money. Contributors read items from a bank of 3,425 prompts covering digit strings, letter-by-letter spellings, personal names and street addresses, one item per recording, each character articulated separately. Reviewers send back rushed readings, wrong sequences and clipped ends; text-to-speech is screened out. Every item is prompted, so nobody reads their own details, and each clip maps to its bank item, so reference text comes with the audio. Speakers come from Kled's global base, Southeast Asian, American and European English foremost.