Just to expand on the point about sample size. There is a tendency to reason as follows:

Wow! I got this interesting result with a small sample. Imagine what it would be like in a bigger (ie better sample)!

Problem when you get that bigger sample the effect/result disappears or is much smaller. When you have a small sample yes it is “easier” to miss things. It is also “easier” for unusual things to show up too.

We see this all the time in epidemiology. You’ll get a study saying eat X and you double your risk of cancer!!!1!!!11!! No causal pathway is suggested. Or they grab multiple possible pathways (it could be this, it could be that, or this entirely different pathway!!!). Then another study comes along saying eat X it is good for you. And all these are are correlations. Usually with no attempt to control for confounding factors (eg Smoking).

This usually the case when the effect you are looking at is also small. For example a type of cancer may have an incident rate of 1% in the population. Finding something that has an effect on that is also small relative to the population. So the power of your statistical tests will also be low. You can try to improve things by increasing your sample size (eg CCP Rise using 80,000 accounts which is more than 2 orders of magnitude higher than the number of accounts here).

My point is absolutely look at the data but keep in mind things like sample size, effect size and statistical power. And no, having lots and lots of data is not necessarily better. If you have lots of observations with lots of variables the finding significant correlations with standard test levels (eg 5%) is “too easy”.