One falling return can have several causes
A paid social channel can become less productive because it reaches the same people more often, creative becomes less persuasive, auction prices rise or the audience mix changes. A single spend-response curve may absorb several of those mechanisms. Calling the curve saturation does not prove that saturation caused the decline.
- Check delivery costs
- Map creative changes
- Challenge the response curve
- Test a controlled refresh
Record spend and delivery alongside campaign structure, major creative launches and measurement changes. Use the saturation guide for the mathematical shape, then ask whether the shape is stable across the operating conditions in your data.
The response you infer depends on transformation choices and what the data can separate. The how marketing mix models work learning path connects this idea to the assumptions, practical checks and next steps.
To work through the connected decisions, read MMM saturation curves: know when you are extrapolating and YouTube MMM data audit: reconcile reach and frequency.
Separate cost inflation from exposure response
A fictional campaign spends £10,000 at a £10 CPM and buys one million impressions. Later it spends £15,000 at a £15 CPM and still buys one million impressions. Spend increased by 50%, but exposure did not. A spend-only response can look flat even if the response to exposure is unchanged.
This example does not establish that impressions cause sales or that all impressions are equivalent. It shows why the units matter. If you model exposure, you still need a defensible cost assumption to convert an optimised exposure level back into budget.
Do not give creative a false causal label
A creative refresh often happens because performance has already fallen. Comparing before and after can mix the effect of the new creative with the reason it was introduced. The refresh may also coincide with a new offer, audience or landing page. Keep a change log rather than treating a creative indicator as automatic causal evidence.
Where feasible, compare creatives under a design that controls the assignment and delivery conditions. For MMM, investigate whether different creative periods require a different response assumption, but resist adding so many parameters that each period explains itself.
Use diagnostics to choose the next experiment
Compare model residuals and response estimates across creative eras, spend bands and auction-cost regimes. If the model's recommendation depends on a small number of launch weeks, show that sensitivity. A channel-level curve may be useful for planning while remaining too coarse to choose individual adverts.
Suppose the proposed action is to increase spend by 20%. First check whether the extra budget buys incremental reach or mostly frequency, whether the creative can support the same response, and whether the auction-cost assumption is reasonable. The validation guide helps separate forecast checks from causal claims.
Choose a bounded budget or creative test with a clear outcome and duration. Keep the result separate from routine platform attribution. The best follow-up may be new evidence about the mechanism, rather than another fitted saturation curve.
Sources and further practice
This article’s worked example is illustrative, not a measured client result. Use the linked methodology to check the assumptions before applying it.