Definition
A process shows diminishing returns when its output grows less than proportionally as input rises. In marketing, the input is usually spend and the output is conversions, revenue or reach. Doubling the budget rarely doubles the result; it adds something, but less than the first half delivered.
The shape is often called a response curve or saturation curve. It is steep at low spend, where the most responsive people and cheapest opportunities are bought first, and flattens as spend rises, where the remaining opportunities are costlier or less likely to respond. Diminishing returns is the name for that pattern; Marginal ROASExtra revenue returned by the last increment of ad spend. is the measure that quantifies it, comparing the change in revenue with the change in spend for the latest increment rather than the total.
In practice
The curve has different causes in different channels, but they share a logic.
- Finite search demand. Only so many people search for a product in a period. Once a campaign is visible on most relevant searches, more budget can only buy higher positions, broader matches or weaker queries, each costing more per result. The bidding hub lists rising cost after a budget increase as a sign of exactly this, to be read through marginal rather than average cost.
- Audience saturation. A finite audience on a social or display platform is reached quickly. Further spend shows ads to the same people again, raising FrequencyAverage times a person saw an ad., or pushes out to less responsive people.
- Auction depth. Platforms deliver the cheapest available results first; bigger budgets reach into costlier auctions.
- Organic effort. The same shape applies beyond paid media. The app store playbook notes that store optimisation work keeps paying, but less, once core terms rank and the main visuals are right.
Seeing the curve
Diminishing returns are rarely visible in a single report, because reports show averages. Two methods reveal the curve. Budget tests change spend in steps and read the incremental result after each, which gives a marginal cost or return for each increment; the testing and scaling guide sets out guardrails for deciding when a step has gone far enough. Modelling estimates the curve from history: a MMMStatistical model of channel impact. fits a response curve to each channel using spend and outcomes over time, as the marketing mix modelling basics guide explains. Click paths alone cannot show it, which is one reason the comparison of MMM and multi-touch attribution treats saturation as something only modelling or experiments can measure.
Average versus marginal results
The average result of a channel is total output divided by total spend. The marginal result is what the last increment of spend bought. Under diminishing returns the marginal result is always worse than the average, and the gap widens as spend grows. A channel can report a healthy average CACCost to acquire one new customer. while its last tranche of budget acquires customers at a loss. Budget decisions are decisions about the next increment, so they should be made on the marginal figure. The guide to marginal returns and forecasting shows how to estimate marginal cost without a full model, and the marketing budget calculator helps set the total budget that those marginal judgements then divide between channels.
A plant nursery finds the edge of local demand
Greenshank Plant Nursery spends £2,000 a month on spring search ads and gets 160 orders, a CPA of £12.50. It doubles spend to £4,000 for April and gets 215 orders. The average CPA looks acceptable at £18.60, but the second £2,000 bought only 55 extra orders, a marginal CPA of about £36, more than the £22 an order contributes. The extra clicks came from broader queries and from shoppers too far away to collect plants. The nursery returns spring spend to £2,400, where the last increment still pays, and moves the rest to autumn bulb campaigns, which reach a different, unsaturated pool of demand.
Related terms
Marginal ROASExtra revenue returned by the last increment of ad spend. is how diminishing returns are measured in revenue terms. A MMMStatistical model of channel impact. estimates the response curve for each channel. FrequencyAverage times a person saw an ad. is the mechanism by which audiences saturate, and CACCost to acquire one new customer. is the figure whose average hides the rising cost of the last customers acquired.
Common confusions
- Diminishing is not negative. Returns still rise with spend; they rise more slowly. Losses begin only when the marginal result falls below what a sale is worth.
- Average performance does not show the curve. A good average can conceal a poor margin.
- A temporary dip is not saturation. Platforms re-learning after a budget change can show short-term cost rises that are not the curve itself.
- The curve is not fixed. New audiences, creative, offers or seasons can shift it outwards.
- Cutting spend is not the only response. Moving budget to a channel or season further down its own curve often buys more than cutting.
