Use Bernoulli for one binary trial, binomial for the number of successes across fixed trials, and normal for many continuous measurements or approximations.

These distributions model different kinds of uncertainty. Interviewers often expect you to connect the distribution to the data-generating process rather than memorize formulas.