Not strictly related to this discussion, but perhaps worth sharing.
I worked at a smaller tech company, and we kept saying "Correlation does not imply causation", and trying to come up with ways to separate them, etc. We spent/wasted a lot of time on this, explaining it to non-mathy people (which leads to even less confidence in data).
Then I went to work at one of the big tech companies, which is probably one of the best in the world at testing/data for product, and there I never heard it. If there was correlation, unless it was clearly spurious or a statistical fluke, they pushed on it hard. "It looks like more X means more DAUs. So let's manufacture more X." And it works pretty well for that company.
My takeaway is, I stopped thinking about this too much. Today, if there's correlation, unless it's clearly spurious/etc, and unless somebody has a better idea, I tell the product/eng team to push hard on it and don't even mention correlation vs causation.
I worked at a smaller tech company, and we kept saying "Correlation does not imply causation", and trying to come up with ways to separate them, etc. We spent/wasted a lot of time on this, explaining it to non-mathy people (which leads to even less confidence in data).
Then I went to work at one of the big tech companies, which is probably one of the best in the world at testing/data for product, and there I never heard it. If there was correlation, unless it was clearly spurious or a statistical fluke, they pushed on it hard. "It looks like more X means more DAUs. So let's manufacture more X." And it works pretty well for that company.
My takeaway is, I stopped thinking about this too much. Today, if there's correlation, unless it's clearly spurious/etc, and unless somebody has a better idea, I tell the product/eng team to push hard on it and don't even mention correlation vs causation.