Just a few minuets ago, I read this: The reason macroeconomics does'nt work too well - Noahpinion's blog. With the eye of a quant finance person, I simplify to two problems: uninformative data and the difficulty to decide which competing model does explain the data best.
Let me take this and come back to the model problems I discussed in recent posts, like When Three Rights Make A Wrong. It is vital that we chose the right model and data to calibrate the model on.
If interest rates can become negative you might need a normally distributed short rate model at hand and data for calibration that fit to that model.
Or, if you want to calibrate, say, a Heston model to price an exotic option and have only market data from vanilla options, calibrating the fits might be very well, but the price calculated by the calibrated model gets wrong ....
Not surprisingly, problems are quite similar in different topics.