You should mostly ignore Media Mix Modeling (MMM), because the models can be dangerously misleading on their own. The core limitation is that MMM is a statistical exercise in correlation, not a measurement of causation. As Brett Curry points out on eCommerce Evolution, these tools are fundamentally “correlation based.” They look at top-line data and see that when spend on a certain platform goes up or down, total sales tend to move in a similar direction. This doesn't prove that the ad spend caused the sales lift. It just shows that two trend lines moved at the same time, which can happen for dozens of reasons.
The one nuance worth knowing is that MMM can become useful, but only when you inform it with causal data. Tom Leonard made a great point on eCommerce Evolution that the real benefit comes from using incrementality testing to feed the model. When you run your own lift tests, you generate actual causal data about how your marketing efforts are working. By providing that data to an MMM tool, you can give the model better constraints and make its statistical correlations more grounded in reality. Without that step, you're just looking at high-level trends that can easily lead you to the wrong conclusions about your ad spend.