Prevention to 100% would be costly. They have diminishing marginal returns.

Why would someone pay $50,000/year for a hypothetical system which would theoretically catch everyone and fill out the police report automatically when the store experiences a tiny $300 annual shrinkage loss?

Because stopping retail theft 100% requires an inconvenience to normal customers that is off-putting and will cause them to go somewhere else.

I certainly wouldn't shop somewhere that required my bags and pockets to be searched every time I left the store.

Lying for Money makes the case that the optimal amount of fraud in any real world system is non-zero. Fraud prevention has direct and indirect costs. At some point the costs of fraud prevention will exceed the cost of fraud itself.