Hackathon guide: setup and judging
A hackathon is a time-boxed build event where teams ship a working prototype against a brief, then a panel judges it. Running one is two jobs: getting good projects built, and judging them fairly. This guide walks through the planning steps in order, then the judging decisions that decide whether your results hold up — the same decisions GavelLab was built around for hackathons, pitch competitions and innovation programmes.
Part 1: Setup and planning
1. Set the goal and theme
Decide what the hackathon should produce — prototypes for a specific problem, learning outcomes for a cohort, or recruitment signal. Write it in one sentence and use it to judge every later decision. A narrow theme makes submissions comparable and judging far quicker.
2. Fix the format and dates
Choose in-person, online or hybrid, then lock the start, submission deadline, judging window and results announcement. Publish all four dates at launch. Leave at least 24 hours between the submission deadline and the results, so judges are not scoring under pressure.
3. Set the rules and eligibility
State team size, who can enter, what participants may build in advance, which licences are allowed, and how AI tools may be used. Ambiguous rules become disputes at judging time, so publish them alongside registration.
4. Budget, venue and prizes
Cost out venue or platform, connectivity, food, mentor time and prizes. Prizes should reward the outcomes named in your goal — a best-first-time-team prize signals something very different from a single grand prize.
5. Recruit mentors and judges early
Invite judges before you promote the event, not after. Aim for a mix of technical and product perspectives, and confirm how many hours each judge can give — that number decides how many submissions each of them can realistically score.
6. Open a single submission channel
Collect every project through one form with the same required fields: title, team, description, repository link, demo video and deck. One channel means no lost entries and no manual copying into spreadsheets before judging.
7. Plan the run of the day
Publish a schedule: kick-off, mentoring checkpoints, submission cut-off, demos, judging and announcement. Checkpoints catch teams who have drifted off-brief while there is still time to correct.
Part 2: Judging that holds up
Most hackathon complaints are about judging, not the build. These steps remove the usual causes: vague criteria, inconsistent scales and invisible progress.
Design the rubric before judging opens
Four to six criteria is the sweet spot — for example originality, technical feasibility, user experience, and impact. Define each one in a sentence so two judges reading it reach the same conclusion.
Weight the criteria deliberately
Weights are where your goal becomes maths. A learning-focused hackathon might weight originality and user experience highly; a client-brief event weights feasibility and impact. Publish the weights with the rules.
Use a fixed scale on every criterion
A 1–5 scale with short descriptions for each point reduces drift between judges. Avoid open-ended totals — they make scores impossible to compare across a panel.
Run rounds instead of one giant pass
A screening round on completeness, then a scoring round, then a final round for the shortlist. Each round can carry its own criteria, so judges spend their deepest attention on the strongest entries.
Turn on blind judging
Hiding team and submitter names while scoring removes the pull of familiar names and reputations. Reveal identities only when standings are final.
Require written feedback
A score without a comment teaches a team nothing. Ask each judge for a short note per submission — it also gives you a defensible record if a result is questioned.
Agree how ties break
Decide in advance: highest score on the heaviest criterion, or a short panel discussion. Deciding after the fact always looks arbitrary.
Part 3: Running a judging portal
However you collect scores, a judging portal should make the panel's work obvious and the results reproducible.
- Give every judge their own account instead of a shared login, so scores and comments are attributable.
- Let judges score from a queue with a Save & Next action — they should never hunt for the submission they have not reached.
- Show progress per judge, so you can nudge the panel before the deadline rather than after it.
- Autosave feedback as judges type; judging sessions get interrupted.
- Keep private notes separate from feedback that participants may see.
- Compute weighted totals and ranks automatically — manual spreadsheet maths is where results go wrong.
- Share final scores, ranks and feedback back to each judge so the panel can sanity-check standings together.
Run your judging on GavelLab
Create your event and rounds, invite judges by email, and let them score every submission against a weighted rubric with blind judging on.
