Four criteria, ten points each, forty points total. The same sheet from the first heat to the grand finale. Panels are defined by roles, not by famous names.
Every judge scores each contestant independently. Scores are integers from 1 to 10 on each criterion. We do not use half points; if you are between two bands, pick the lower one unless the talk clearly earned the higher. A judge’s sheet therefore totals 4–40. The heat ranking is the mean of the sitting panel after a single recusal adjustment (see below). Video screening uses the same four columns, scored on the 90-second file rather than the live seven minutes.
| Criterion | What judges are watching for | Points |
|---|---|---|
| Comprehension | Did the talk make a complicated idea legible to a non-specialist? Could a smart stranger summarise the work back in one sentence? | 1–10 · 25% |
| Content | Was the underlying work substantive? Is the problem real, the approach sound, the evidence specific — not a slogan? | 1–10 · 25% |
| Engagement | Did the speaker hold the room? Did they earn attention, or rely on the format to keep it? | 1–10 · 25% |
| Communication | Were pacing, language, slide, and stage presence working together — or fighting each other? | 1–10 · 25% |
The idea is not recoverable. Jargon without a translation, no problem, or a slide that is a paragraph. A specialist in the field still could not retell it.
A competent seminar, not a 7MT talk. The problem is named but not felt. Evidence is vague. The room is polite rather than leaning in.
A non-specialist can retell the work. One image holds. Timing is inside the bell. Something still generic — the analogy is borrowed, or the last minute is a résumé.
The room could quote a sentence on the way out. Mechanism and evidence are specific. Eight is rare; nine is a finale talk; ten is almost unused and must be defensible in closed session.
A disqualification (time, props, second slide) records a zero for the round and is not averaged. Ties on mean score are broken in closed session: first by content, then by comprehension, then by a show of hands. Audience votes do not break a judges’ tie.
Every level of the competition also carries a People’s Choice award decided entirely by audience voting. It is not a tiebreaker and not a fourth prize that can overrule the panel. It is its own certificate. Many finalists value it most. Heat audiences vote on paper or a local form; finale audiences vote online during a locked window after the last talk.
We do not publish a roster of famous individuals as if they had already confirmed. Panels are seated by role so a speaker from any background meets a comparable mix. Names for a given heat go on the door the morning of the event. The 2026 grand finale panel is announced in October.
Every regional heat seats three to five judges. The seat plan is:
Panels are rotated annually and balanced, as far as a small city allows, for seniority and discipline. Judges declare conflicts (advisor, co-founder, family, current employer of the speaker). A declared conflict does not empty the chair — recusal of a whole judge would shrink the panel and reward whoever sat next to a conflict. Instead the conflicted judge still scores; the producer discards that one sheet for that one speaker and renormalises the mean. If two judges conflict on the same speaker, the producer seats a standby for that talk only.
Print the rubric. Have three colleagues sit as specialist, operator, and communicator. Time the talk. If they cannot retell the mechanism, the content column will not save you. Request the PDF scoring sheet from Resources. Then apply: Apply.
See past finalists