> I am not sure what you mean by that. How is supply-demand model better? What it gives you in those situations?

It yields better predictions with fewer inputs. For simple calculations of optimal pricing or optimal production quantities given input costs, a simple table of volume versus price beats a matrix. IO models become interesting when observing entire industries. (That is why the DoJ uses the HH metric [1], an IO-derived metric, to measure competitiveness.) But they're macroeconomic, not microeconomic, models. Macro has a worse track record than micro, which starts with simple regressions of volume and price that we call supply and demand.

These kinds of matrix models are deeply embedded into the IMF's approach, by the way. Many are critical of the IMF's over reliance on such difficult-to-intuit models.

[1] https://en.m.wikipedia.org/wiki/Herfindahl_index

Again, I think you're talking about statistical model, which I don't object to. I object to supply-demand model as a phenomenological model, as taught in the economics textbooks. What they are doing is not regression.

Basic rules like "if you decrease supply the price will go up" are true in more cases than they aren't. I agree, however, that entry-level students tend to overfit their models. If they never receive later training, they go into the world with a faulty view. Those same people then find those models failing in places we've known, for decades, they fail (in some cases, in a predictable way) and then come back to blame the models.