A retention budget lets you act on a limited number of employees. The instinct is to target whoever's most likely to leave. But the people most likely to leave often aren't the people you can change. Pick a strategy and watch the same budget save very different numbers.
The outcome is expected stays you caused — the sum of each treated person's uplift (how much your action lowers their chance of leaving). Targeting by flight risk pours the budget into lost causes who leave anyway; targeting by uplift finds the persuadables, the only group whose decision your action actually changes. The gap is widest when the budget is tight, which is when it always is. Predicting who leaves is not the same as knowing who to keep — that second question is causal, and it's the one a flight-risk model can't answer.