Diminishing Returns: When More Spend Stops Buying More Sales

There's a particular kind of tiredness that sets in at a party well after most of the room has already had a good time. The host keeps offering the same story to anyone still listening, and the story hasn't changed, only the size of the audience left willing to laugh at it. An ad campaign gets tired the same way. The first pound of spend on a channel reaches the people most ready to respond. Every pound after that reaches someone slightly less ready, until a channel is mostly refilling the glasses of people who'd already said yes.

Marketing mix modelling calls this diminishing returns, or saturation: as spend on a channel rises, each additional pound produces a smaller incremental effect than the pound before it. Google's own documentation for its MMM tool Meridian describes it as the "intuitive principle that as spending... increases, you eventually see diminishing marginal returns" (Media saturation and lagging, Meridian). The response curve bends, rather than continuing in a straight line, because the audience with genuine appetite left to respond is running out.

Average ROI flatters a channel; marginal ROI tells the truth

A channel's headline return on investment is an average across every pound spent on it, including the very first pound, which almost always performs best. That average can look healthy long after the channel has stopped being a good place to spend the next pound. For example, a paid search channel running at 87% of its estimated saturation, delivering a marginal ROI of 1.6x, sitting next to a TV channel at only 35% saturation delivering 3.7x on the next pound. Shifting budget from the former to the latter unlocks millions in revenue.

The practical read is the equimarginal principle: budget is allocated well when the marginal return on the next pound is roughly equal across channels, not when each channel's average return looks respectable in isolation. A channel with declining conversion despite steady or improving cost-per-click, and rising acquisition cost despite steady ROAS, is often not becoming less efficient. It's saturating, which shows up as exactly that pattern of headline metrics holding while marginal performance quietly erodes.

A saturation curve is only trustworthy within the range you've actually tested

Google's Meridian and Meta's Robyn both fit a curve to describe how a channel's returns bend as spend rises, using your actual spending history to estimate where that bend happens for each channel. That's exactly why the curve is only as good as the data it's built from: it's a reasonable description of how a channel behaves within the range of spend you've actually run, not a reliable forecast of what happens well beyond it.

If a channel has never been pushed particularly hard, the model is effectively guessing at a ceiling it's never seen, and that guess can overstate how much headroom is really left. Before doubling down on a channel because "the model says there's room," it's worth asking whether that room has ever actually been tested, or whether the model is extrapolating into territory it has no real evidence about.

The half-saturation point is the parameter worth the most scrutiny before trusting a curve. Google's own research paper introducing this approach flags that it's poorly identified when the actual observed spend never gets close to it. The model is effectively extrapolating a ceiling it's never seen, which risks overstating how much room a channel has left (Bayesian Methods for Media Mix Modeling with Carryover and Shape Effects, Jin et al., Google). A saturation curve fitted on a narrow spend range is a reasonable estimate of behaviour within that range, not a reliable forecast of what happens well outside it.

Reallocating budget only moves the problem; better creative removes it

Once a channel is identified as saturated, the reflexive fix is to shift budget toward a channel with more headroom, which is the right first move, following the equimarginal logic above. But it treats the ceiling as fixed, when a large part of that ceiling is set by how good and how varied the creative running through the channel actually is. Meta's own internal analysis of ad exposure found that conversion likelihood falls in a predictable, near-monotonic pattern as the same user sees an ad repeatedly, roughly a 45% lower likelihood of conversion by the fourth exposure, and that refreshing the creative on fatigued ad sets produced a measurable, dose-dependent recovery, on the order of an 8% lift in conversion rate in the most fatigued cases (Creative Fatigue, Analytics at Meta).

Nielsen's research points at the same lever from a different angle: raw share of voice (how much a brand spends relative to competitors) and effective share of voice (adjusted for how good the creative actually is) can diverge by as much as 8 percentage points, and buying that same gap through spend alone cost major advertisers an average of £1.5m in the market Nielsen studied (How to Increase Share of Voice Without Increasing Media Investment, Nielsen).

The saturation curve isn't only a budget constraint. It's a creative one. A saturated channel with strong, varied creative reaching fresh audiences will bend later than the same channel running one tired asset on repeat, which is the more durable fix once the easy budget shifts have already been made.

If it's been a while since anyone checked which of your channels are still earning their next pound and which are just topping up an audience that's already said yes, that's usually worth a proper look. Get in touch with Day1data to see where your own spend actually sits on the curve.

FAQ

What are diminishing returns in marketing mix modelling? The pattern where each additional pound spent on a channel produces a smaller incremental effect than the pound before it, because the audience most ready to respond is reached first and everyone after that is progressively less receptive.

What's the difference between average ROI and marginal ROI? Average ROI blends the performance of every pound ever spent on a channel, including the highly efficient early spend. Marginal ROI is the return on the next pound only, the number that actually matters for a budget decision.

How does an MMM model diminishing returns? By fitting a curve to your actual spend and outcome history for each channel, describing how sharply returns bend as spend rises. The curve is only as trustworthy as the range of spend it was built from.

What does it mean if a channel is "saturated"? Its current spend is far enough up its response curve that the next pound spent there returns meaningfully less than a pound spent on a less-saturated channel.

Should a saturated channel's budget always be cut? Not necessarily first. Reallocating to a less-saturated channel is the fastest fix, but improving or refreshing the creative running in the saturated channel can raise the ceiling itself, which is often more durable than simply moving the same budget elsewhere.

Does more frequency always mean fatigue? No, but it trends that way for direct-response objectives specifically. Conversion likelihood tends to fall in a fairly predictable pattern as the same person sees the same ad repeatedly, which is exactly what varied, refreshed creative counteracts.

How reliable is a saturation curve if a channel's spend has never been pushed very high? Less reliable. The curve is fitted from your actual spend history, so estimating where it bends beyond spend levels you've actually tried is closer to a guess than a measurement. Treat a curve built on a narrow spend range as directional, not exact.

How does saturation interact with adstock? They're separate mechanics measuring different things. Adstock describes how long an effect lasts over time, saturation describes how the size of that effect changes with spend, but a channel's real response curve is shaped by both together.