Why does my best channel get worse as soon as I spend more on it?
Because you judged it on average performance and you buy at the margin. Each additional increment of spend reaches less-ready demand than the spend already in place, so the incremental leads convert worse than the channel average. A channel can remain the best on average while its marginal return has already fallen below another channel's. Budget decisions should be made on marginal returns, not blended averages.
The average includes demand you would have gotten anyway
A mature channel's average performance is dominated by its cheapest, highest-intent inventory: people searching for exactly what you do, at the moment they need it. That inventory is finite. Once you are capturing most of it, additional budget necessarily goes to broader terms, wider geography, colder audiences or worse positions.
So the average stays attractive while the next increment is not. Scaling on the strength of the average is how a channel that looked excellent turns mediocre at higher spend, and why the person who recommended scaling it gets blamed for something that was arithmetically inevitable.
How to actually observe the marginal number
You cannot read marginal return off a standard report. You have to create a change and measure the delta. Move one channel's budget in a meaningful step, hold everything else steady, wait long enough for the revenue to mature, then compare the change in booked jobs to the change in spend.
That comparison is the marginal cost per booked job. Do it once per channel per season and you will have a rough returns curve. It is coarse, and it beats every model that infers marginal performance from averages.
- Change one thing. Simultaneous changes across channels make the deltas uninterpretable.
- Wait for maturation. Reading the delta before the revenue lands understates the increment.
- Watch capacity. If the extra demand hit a full schedule, you measured your capacity, not the channel.
- Note the season. A step change during a seasonal ramp will be credited to the budget change. Compare against the prior year's shape.
Where the curve actually bends
Impression share and lost-impression-share-to-budget are useful early indicators of headroom in search channels. When you already hold most of the available impression share for your core terms, the remaining growth in that channel is coming from broader matching, and broader matching is where revenue per lead falls.
The same logic applies to Local Services Ads and to Meta, though the constraint shows up differently: in one it is available demand, in the other it is audience saturation and frequency. Both bend; they just bend on different axes.
What to do with the finding
When a channel's marginal return drops below another channel's marginal return, move budget. Not because the first channel is bad, but because the next unit of money buys more elsewhere. That is the entire decision rule, and it is why blended cost per lead is a poor allocation input.
It also means the answer changes with the season and with your capacity, so the exercise is periodic rather than one-time. Tying spend decisions to booked and completed jobs rather than platform conversions is the point of connecting paid media to the operational record.
Topics: marginal return · budget · paid media · revenue per lead · diminishing returns
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.