The biggest channel already reads as close to fully covered, taking almost all of the clicks in its category, while the new-account target stays fixed regardless. Growth is supposed to come from channels the model cannot yet size with confidence, wide intervals on numbers everyone has been told to treat with caution rather than certainty. Testing that means moving money out of the one channel the whole business already credits for its results, on evidence not yet strong enough to protect the case if the move is wrong. If new accounts dip after the budget shifts, the decision carries a name and no defensible number behind it yet.
A right-sized read shows how much more the dominant channel can actually absorb before it stops paying back. That answer arrives before real budget moves, not after. A number for the channels waiting on funding replaces a wide interval nobody can act on.
The channel carrying most of the volume is expected to hit a ceiling eventually, but nothing currently locates where that ceiling actually sits. When growth flattens for a quarter, a real limit on the channel and a slip in execution look identical from the outside, and the two call for opposite responses. Either the wall arrives with no warning and derails a plan built on the old growth rate, or the blame lands on a plateau that was there to be found the whole time.
A read on the channel's own history shows where returns start to bend, rather than a guess about how much room is left. That distinguishes a structural ceiling from an off quarter before next year's plan commits to either explanation. And the finding lands early enough to plan around instead of surfacing after the fact.
Saturation is the point where a channel already carrying most of total volume stops returning proportionally more for additional spend. Beyond that point the marginal cost of the next customer rises, even while the channel's average numbers still look healthy, because an average blends the productive early spend with the unproductive spend layered on top of it.
The channel's own spend history is read for the point where returns start to bend, using regional or time-based variation already present in that history rather than a new experiment imposed on it. Where that variation is not enough on its own, a bounded, right-sized test on the channel fills the gap.
The logic holds regardless of which channel dominates: the model reads that channel's own spend and outcome history for the point where marginal returns bend, rather than assuming a fixed share of clicks means the channel is either exhausted or wide open. A channel covering most of a funnel is exactly the case this is built for.
When the dominant channel does not have enough history, or enough variation in its spend, to separate a real ceiling from an ordinary slow quarter. In that case the honest answer is that the ceiling cannot yet be located, not a confident-sounding estimate.
Because average return and marginal return are different numbers, and reporting shows only the first one. A channel can keep returning an acceptable blended figure long after the return on each additional unit of spend has started to fall, so nothing in the dashboard flags the moment it passed its efficient ceiling. Whether a given channel has passed it is not a question a benchmark answers: the read comes from that channel's own spend and outcome history, at the point where the curve bends.