Test It with 30 People
Inspired by Crystal W episode
Crystal Widjaja ran experiments at Gojek with sample sizes of 30. The precision changes, but the direction doesn't.
Crystal Widjaja says run the experiment even if you only have 30 users. The data you get back won't change much at larger scales. What's better than having 30 data points? Having 100. But what's better than having zero is definitely 30.
You don't need a control group to test whether your toddler eats more vegetables when you call them "dinosaur trees." Just try it. See what happens. You're not publishing in Nature. You're trying to get spinach into a small human.
The same toddler who needs "scientific validation" for vegetables will eat a crayon off the floor without a second thought. No A/B testing. No user research. Just pure commitment to chaos.
So yeah, call the broccoli "trees for tiny dinosaurs." If it works once, it's data. If it works twice, it's a pattern. If it doesn't work, at least you didn't waste six months planning the perfect experimental conditions while your kid survived entirely on goldfish crackers and spite.
