But those metrics are hard to move, so you target secondary metrics.

The problem with that strategy becomes obvious when you spell out the consequences: measurably improving the product is hard, so you measure something else and hope you get product improvements.