<b>Why does a channel's measured ROI fall the moment you act on it?</b>

<b>Why does a channel's measured ROI fall the moment you act on it?</b>

A frustrating and predictable pattern: you find a winning channel, scale it, and its efficiency collapses. This is saturation, and misreading it leads to chronic misallocation.

<b>The mechanism</b>

Marketing response is non-linear. Each channel follows a <i>saturation curve</i> — strong returns at low spend, diminishing returns as you scale, because you exhaust the cheap, responsive audience first. The average ROAS you measured at $10k/month does not hold at $100k/month. You were reading a point on a curve and treating it as a constant.

<b>Why attribution hides it</b>

Most attribution and ROAS reporting is <i>average</i>, not <i>marginal</i>. Average ROAS tells you the blended return on all spend; the decision you actually face — should I add the next dollar? — depends on <i>marginal</i> ROAS, which is always lower and falls as you scale. Optimizing on average ROAS systematically overspends into the flat part of the curve.

<b>The nuance</b>

Saturation is also why incrementality results expire. A channel measured as highly incremental at low spend can be near-saturated after you scale it, so last quarter's lift factor over-promises today. Effects are conditional on spend level, not fixed channel properties.

<b>What to actually do</b>

— Make decisions on marginal return, not average ROAS. The question is the slope, not the level.
— Estimate response curves (MMM does this explicitly) so you can find the point where marginal return crosses your hurdle rate.
— Re-measure incrementality after any meaningful scale-up; lift factors don't survive a 5x budget change.

Bottom line for practitioners: a channel's ROI is a curve, not a number, and you slide down it the moment you scale. Budget on the margin and treat every lift factor as valid only at the spend level where you measured it.