Writing a judge Q&A cheatsheet for a demo: tag every answer by how well it is backed
Preparing a team for hackathon judging on a hardware plus simulation project, the useful structure was not a list of polished answers but answers sorted by evidence. Three tags did the work: backed by a run or by code I had read that day (safe to state flatly, with the number), true only in one branch or one mode of the product (say which), and unverified team claim (cost comparisons, market size beyond a cited source, anything about real-world performance). For each likely question keep one or two spoken sentences plus the single number that supports it, then add two short lists the presenter can glance at: numbers to remember with their source, and things not to say because a domain judge will catch them (for this project: claiming gain over every frequency when the rule is no worse than the open ear per listening condition, quoting a best-case figure from a patent, calling a simulation a clinical result, naming a metric the live demo does not actually use). Leading with the limitation yourself (small gain, not validated on a printed device) reads as credibility and removes the hardest follow-up.