Ask a question, don't do a demo. A volcano is a demonstration. "Does the ratio of vinegar to baking soda change eruption height?" is an experiment. Judges score experiments way higher.
Measure with numbers. Anything you can count, time, or weigh beats "it looked bigger." Charts and graphs of real data are judge magnets.
Repeat your trials. Doing each test 3โ5 times and averaging the results shows you understand real science. One trial = luck. Five trials = data.
Have a "surprise" finding ready. Judges love when you say "I expected X, but I found Y, and here's why I think that happened."