Making data driven decions
Decision theory says to us, there are two major scientific ways to arrive at decisions, especially where causes of issues aren't scientifically decoded yet. These are
Hypothesis led (which I will call H-decisions)
Data driven (… D-decisions)
H-decisions have been quite prevalent, you start with a short hypothesis, setup a falsifiability criteria, and then go out testing it. We have done it a thousand times in our life. When a bulb goes out, we instantly make a hypothesis -- maybe the filament/ led is blown. More often than not, we check it out. If the bulb glows when put in another socket, we know our hypothesis is wrong, if not, we are perhaps correct.
Two things are really important here
Falsifiability - the fact that bulb working in another socket is a sure defiance of our initial assumption
Non confirmation - as an irritating philosopher, one may always pose that, to completely prove my fused bulb hypothesis, I have to test infinitely many sockets
The cheat code to both of these above is statistics. Elaborate testing methods, which make sure that one can bring certain confidence that an hypothesis is indeed not false. (My stats prof hammered this non false thing, so..)
Now, why am I saying all this. Because data and statistics often happen to be corner stone for D-decision making. So, often when we go out with a hypothesis, and back it up with data to prove/ disprove, we pat ourselves thinking we were data driven. But a subtle line of difference exist between the two, and that is what unlocks the true powers of each.
D-decisions, don't necessarily look for a reason. When a strong correlation is found, D-decision is to work on it thought asking why. It is indeed scary, and that is why data truthfulness matters
D-decisions, are useful, only when data can be synthesized. If one were to look at data and take decisions, it would be much of a H-decision. A true D-decision, needs us to find combinations, and cross relations between them
In our bulb case, a D-decision would come, if we observed several bulbs for there lifetimes, and then made a model which predicts the conditions in which a bulb blows out.