What is causal inference?

The alure of machine learning is strong with complex models that excel in predicting with high precision, as well as in categorizing or grouping your data. However, in diagnostic analytics, you are not forecasting—you’re trying to understand why something happened. You are trying to understand the drivers of an outcome. And this is a fundamentally different question. It is not just semantics or academic pedantry. Forecasting something and understanding the effects of something are two different tasks in statistics.

Without a causal understanding of the world, it’s often impossible to identify which actions lead to a desired outcome. In business, we are constantly taking actions to achieve a certain outcome (for example, increase sales). So, in order not to waste our time heating our proverbial thermometer, we need a solid understanding of the causal relationships underlying our business processes. This is the premise of decision intelligence...