Every channel has diminishing returns. Set a curve for each, then watch the optimizer split a fixed budget to squeeze out the most subscriptions — the exact logic a marketing mix model uses to allocate spend.
Response-curve optimizationMarginal-return allocationIllustrative curves — not benchmarks
Total spend / mo
—
across 5 channels
Subscriptions / mo
—
predicted output
Blended CAC
—
cost per subscription
vs. even split
—
subs from smarter allocation
$1.2M
Optimal mode: drag the total budget and watch where the model puts each dollar. Switch to Manual to set your own split and see how much you leave on the table.
Try a budget:— watch Connected TV switch on only at scale.
Channels
Optimized allocation
Diminishing returns
subs vs spend
Where the money goes
Value of the next $1K
marginal return
The idea in 60 seconds
Stop allocating by average ROAS. Allocate by the next dollar.
The first dollars into a channel buy a lot of output. Later dollars buy less — that bend is the channel's saturation curve. The mistake most teams make is judging a channel by its average return, which stays flattering long after extra spend has stopped paying off.
The fix is to look at the marginal return — what the next dollar earns — and keep shifting budget toward whichever channel's next dollar is worth the most. When the next dollar is worth the same everywhere, you've found the allocation that maximizes total output. That equal-marginal point is what the Optimize button solves for.
1 · Shape each channel. A saturation curve says "this channel can deliver up to X, and it takes Y spend to get most of the way there." Efficient channels bend early; upper-funnel channels stay linear longer.
2 · Read the marginal return. The slope of the curve at your current spend is the value of your next dollar. Steep slope = underfunded. Flat slope = saturated.
3 · Equalize the margins. The optimizer moves budget from flat-slope channels to steep-slope ones until every channel's next dollar is worth the same. That's the math behind "where should the money go?"
Is this a marketing mix model?
No — it's the optimization half of one. A real MMM estimates these curves from your historical data, controlling for seasonality, pricing, and cross-channel halo. Here you set the curves by hand to build intuition for how the allocation behaves. Same logic, illustrative inputs.
Why marginal ROAS instead of average ROAS?
Average ROAS is what a channel earned across all its spend so far; marginal ROAS is what the next dollar will earn. Because of diminishing returns, a channel can have a great average and a terrible margin at the same time. Budget decisions live at the margin.
Why does the optimizer sometimes fund a channel with $0?
If a channel's first dollar is worth less than another channel's last dollar at the current budget, it doesn't earn a place yet. Upper-funnel channels often only "switch on" once the budget is large enough — drag the total up and watch them appear.
This is how I think about budget
I run this math on real budgets for a living
This toy uses curves you set by hand. On a real budget, the curves come from your data — and getting them right is most of the job. If you're building or pressure-testing a growth budget, let's talk.