A training manager rewrote a quiz-generation prompt after complaints that questions were too easy. She changed the audience description, the difficulty instruction, the number of questions and the format all at once. The next quiz was better, and she wants to announce the fix to the team. What should she do first?

Answer: B — Test the new prompt on several varied topics, including a hard one, before sharing it.

One good result can be chance, and changing several things at once means she doesn't know which change mattered. Testing across varied inputs, as in (b), is the evidence needed before telling the team. (a) declares success on a single sample. (c) relies on self-assessment, not results. (d) changes things again before the current fix is understood.