Poisson, Pareto/power/zipf and normal distributions are really important. The top 3 for me. (What am I missing?) And often misused (most often normal). It’s really good to know which to use when
Beta
But it is the right distribution to represent uncertainty about the probability of binary events (eg a website user clicking some button). For example, if I have absolutely no idea the probability then I use the uniform distribution, Beta(1,1), which is the maximum entropy distribution. Then if I observe one user and they happen to click, I have Beta(2,1), and at a glance I known the mean of that (2/3) which is a useful point estimate.