Your blended CAC looks fine. The dollar you are about to spend does not. Why averages hide the cost of scale.你的综合获客成本看着没问题，但你即将花出去的那一块钱并不是。为什么平均值会掩盖放量的真实代价。
Here is the most expensive mistake in scaling: judging a budget increase by your blended CAC. Blended is an average of everything you have already spent. The decision in front of you is not about the average — it is about the next dollar, and the next dollar almost always costs more than the last.放量时最贵的一个错误，是用"综合获客成本"来判断要不要加预算。综合值是你已经花掉的所有钱的平均。而你面前的决策跟平均无关——它关乎"下一块钱"，而下一块钱几乎总是比上一块更贵。
As you scale a campaign, you exhaust your cheapest audiences first. The platform keeps finding people, but each new cohort is a little less likely to convert. Your reported CAC drifts up slowly because it is diluted by all the cheap conversions you already banked. The marginal CAC — the cost of the customers you are buying right now — is already much higher.当你放大一个计划，最便宜的人群会先被消耗掉。平台还能继续找到人，但每一批新人群转化的概率都更低一点。你看到的 CAC 上升得很慢，因为它被你早已收入囊中的廉价转化稀释了。而边际 CAC——你此刻正在购买的那批客户的成本——其实已经高得多。
The practical fix is to watch marginal CAC, not blended, when you make scaling calls. Increase budget in steps, hold for the learning to stabilize, and read the CAC of just the incremental spend. The moment that marginal number crosses your payback threshold, you have found the ceiling — pushing past it burns money even while the blended average still looks healthy.务实的解法是：做放量决策时盯边际 CAC，而不是综合 CAC。分档加预算，等学习稳定，只读"增量那部分支出"的 CAC。一旦这个边际数字越过你的回本阈值，你就找到了天花板——再往上推就是在烧钱，哪怕综合平均还显得很健康。
This is also why a single blended ROAS target is a blunt instrument. The right ceiling is the spend level where marginal return still clears your margin. Find it, sit just below it, and put the rest of your energy into raising the ceiling itself — with better creative and a better offer — rather than bludgeoning a tired audience with more budget.这也是为什么"一个综合 ROAS 目标"是个很钝的工具。正确的天花板，是"边际回报仍能覆盖你毛利"的那个支出水平。找到它、稳稳停在它下方，再把余下的精力用在抬高天花板本身——靠更好的素材和更好的优惠——而不是用更多预算去硬砸一个已经疲劳的人群。