In some situations, "perfect is the enemy of the good" is a reasonable approach. Unfortunately, we're not dealing with a good-but-imperfect filter, but a practically useless filter, one that imposes significant costs on top of being ineffective.
"Weapons of Math Destruction" presents good evidence that 90% is much lower than the actual effectiveness of X.
"Weapons of Math Destruction" presents good evidence that 90% is much lower than the actual effectiveness of X.