Decide if ad load is optimized

Let \(L\) denote the ad load, defined as the number of ads served per 100 content units in the home feed. We want to choose \(L\) to maximize expected long-term business value subject to user-experience guardrails. I’ll assume the current production load is \(L_0\), and we test candidate loads around it. 1. Primary success function and guardrail metrics The primary business objective is a constrained optimization: $$ L^* = \arg\max_{L} \; \text{RPM}(L) $$ subject to hard…