The concept of being "data-driven" is misguided in the tech industry.
People often use the term "data-driven" to emphasize the importance of relying on data rather than conjecture and intuition. The idea is to base decisions on factual information.
However, the approach of being "data-driven" and deriving general principles solely from specific observations is unscientific. It is comparable to inductivism, which was refuted by Karl Popper through two objections.
First, data is not objective. Observation is influenced by preexisting theories, shaping what is measured. Our view of the world is filtered through theories and mental models.
Second, data does not generate theories; it only invalidates them. Theories arise from our creativity: it is our ability to simulate the world in our heads and create an endless stream of new explanations. Intuition is simply the System 1 (from Thinking Fast and Slow) output of our world model.
The origin of the theory does not matter. They can come from a fever dream or an LSD trip (both of which have resulted in ground breaking scientific theories). What matters is that the theory is subject to rigorous testing — that is where data comes in. Data is used to falsify our theories, so that we can come up with better ones.
One of Popper’s core ideas is the asymmetry between proving a theory right vs wrong: it is impossible to prove a theory right (Hume’s problem of induction), but it is possible (sometimes trivial) to prove a theory wrong.