The day · Page 56
16:40, judging
For you only
Judge on signal, not polish. The rubric scores three things, each one to five:
| Criterion | Question | Score |
|---|---|---|
| Served the brief | Did it build for the user and goal we set? | 1 to 5 |
| User signal | What did real users actually do with it? | 1 to 5 |
| Worth building | Is this worth taking forward? | 1 to 5 |
Highest total wins. If two are close, the room talks it out. The judges own the call.
Score the learning, not the looks. A rough demo that taught you something real beats a polished one that did not. Say that sentence to the judges before the first pitch, because without it, humans reliably reward the shiniest screen. The bigger hackathons score the same way: a few criteria, each one a question a judge can answer, aimed at the stage the team is actually at.
Who judges. The sponsor plus one or two others: you, or a senior person from a function that did not build. Keep it odd-numbered if you can. Every judge scores every team on the printed scorecard (Appendix A) during the pitches, not after, while the memory is fresh.
Deliberation is ten minutes, in the corner, not a meeting. Add the scores, surface the spread, talk only about where judges disagree by two points or more.
If two teams genuinely tie, use a silent tie-break so the loudest judge does not win by volume: each judge writes their pick and one sentence why, reveals together, and the sponsor calls it. That is note-and-vote again, and the silence matters for the same reason it did at 11:40. If you want a visual version, dot voting works: NN/g's Sarah Gibbons documents the mechanics, including its known failure mode, people voting where the dots already are, which is exactly why the votes go down before anyone sees anyone else's.
The winner is a direction, not a trophy. The real output of judging is the input to the next ten minutes.