Law of Diminishing Returns

Law 05 / 20 Grow the Brand

As you keep adding spend to the same channel, each extra dollar usually brings fewer additional results, so the next dollar matters more than the average.

  • Economic principle
  • 6 min read
  • Documented study
  • Reviewed by Ali Dezhami, November 8, 2026

The dilemma

Illustrative scenario

Jonah runs paid social for an online bedding brand. At $6,000 a week, the channel has been delivering orders at an average cost the finance team is happy with, and each order earns about $90 in contribution margin.

With a quarterly target slipping, the CEO asks him to double the budget to $12,000 a week. Her reasoning is simple: if $6,000 buys this many orders, $12,000 should buy roughly twice as many, and the average cost per order would still sit comfortably below $90.

Jonah has seen this before. The last time he scaled a lookalike audience, frequency climbed, the same people kept seeing the same video, and each extra thousand dollars brought in noticeably fewer new orders than the one before. The average cost per order still looked fine. The conflicting metric: the average says scale, while the cost of the additional orders may already be higher than what they earn.

The short answer

The law of diminishing returns says that if you keep adding more of one input while other inputs stay fixed, each extra unit eventually adds less output than the one before. It is a principle from classical economics. In marketing it describes how extra spend in one channel, audience or creative tends to buy fewer additional results as the most responsive people are reached and the same people see the message again.

How strong is the evidence?

Economic principle

A general principle of economics. The shape of the curve has to be measured case by case.

This is an economic principle that also shows up in advertising data. Simon and Arndt (1980) reviewed published studies of how sales respond to advertising and concluded that the typical response is concave, with diminishing returns, rather than an S-curve with a threshold that must be crossed before advertising works. Modern media mix models build the idea in directly: Jin and colleagues at Google (2017) model each channel with a shape effect, in which response flattens as spend rises, and a carryover effect, in which advertising keeps working for some time after it runs.

Conditions: where the curve bends depends on audience size, creative fatigue, auction competition, seasonality and the time window. A channel can be saturated with one creative and not with another, or in July but not in November.

Measurement matters: platform attribution often credits sales that would have happened anyway, which can make high spend look more productive than it is and hide saturation.

Open questions: whether some media need a minimum weight before they work at all is still debated, and curves estimated from historical data are uncertain when spend has not varied much. The principle is well established; the precise curve for your channel is always an estimate.

Investigating an example

Documented study

How media mix models represent saturation and carryover

Jin, Wang, Sun, Chan and Koehler describe a Bayesian media mix model in which each channel's spend passes through a saturation curve, so additional spend produces progressively smaller increases in sales, and through a carryover function, so one week's advertising also affects later weeks. They show that these curve parameters can be hard to pin down from typical historical data and that results depend on modeling assumptions and prior information.

Limits: this is a methods paper, and much of its demonstration uses simulated data. It shows how diminishing returns are modeled, not where any real advertiser's curve bends or how fast returns fall. Finding that out still requires your own data, ideally with deliberate variation in spend or controlled experiments.

A published study or record, linked below. Its limits are stated.

Choose your move

You are in the dilemma above. Which move would you make? Open each one to see its trade-offs.

A Double the budget in the same audiences with the same creative.

Total orders will probably rise, but if the channel is near saturation the extra orders will cost far more than the average suggests, and frequency and creative fatigue will climb. It could still be right if the additional orders cost less than they earn, which nobody has checked.

B Raise spend in steps, with fresh creative and new audiences, tracking the cost of each step's additional orders.

Slower to reach the target, but it shows where the curve bends. It needs discipline to stop when the marginal cost passes the margin, and weekly step results are noisy.

C Hold this channel flat and test the extra budget in a less saturated channel.

Spreading spend can lift total return when one channel is saturated. A new channel has its own learning costs, though, and weaker measurement can make the comparison unfair in either direction.

No move wins in every business. The law tells you what to check, not what to choose.

How to apply it

  1. Separate average from marginal. For each budget step, calculate the cost per additional order, not just the overall average cost per order.
  2. Create deliberate variation. Change spend by region or by period in a planned way, so the response curve can actually be estimated.
  3. Count incremental results. Use holdouts or geographic experiments rather than platform-attributed conversions, which can include sales that would have happened anyway.
  4. Set a stopping rule in advance. Stop scaling when the marginal cost per order exceeds contribution margin per order, after returns and discounts.
  5. Re-test regularly. Creative, audiences and seasons move the curve; check again at least each quarter.

What to measure

Incremental orders per budget step, marginal cost per incremental order, frequency per person reached and contribution margin per order. Guardrail: a rising average cost per order is a late signal; act on the marginal figure, and confirm it with an incrementality test rather than platform attribution alone.

Worked example

Worked example: scaling one channel in $2,000 steps (illustrative numbers)
Weekly spendIncremental ordersExtra orders from last stepCost per extra orderAverage cost per order
$2,000 50 50 $40 $40
$4,000 85 35 $57 $47
$6,000 110 25 $80 $55
$8,000 128 18 $111 $63
$10,000 140 12 $167 $71
$12,000 148 8 $250 $81

Illustrative numbers, not data from a real brand. With $90 contribution margin per order, every step beyond $6,000 costs more per extra order than it earns, even though the average cost per order stays below $90 all the way to $12,000. Weekly contribution after ad spend falls from $3,900 at $6,000 to $1,320 at $12,000.

Where it breaks

  • Not every channel is near the bend. A small budget in a large market can be far from saturation, and scaling there may be efficient.
  • Thresholds may exist. Some media or creative may need a minimum weight to register; the evidence is debated.
  • Carryover distorts weekly reads. Effects spill into later weeks, so cutting a channel can look free at first and expensive later.
  • Attribution can hide saturation by crediting conversions that were not caused by the ads.
  • Averages mislead. An acceptable average cost can conceal steps that lose money.
  • Ethics: squeezing out the last conversions with ever-higher frequency turns advertising into pestering; set frequency caps.

What if you apply it incorrectly?

A team learns that returns diminish and cuts every channel back to its cheapest first dollars. The average cost per order looks excellent in the next board deck, but total profit falls, because several channels were still well below the point where an extra order cost more than it earned. Diminishing returns tells you where to stop, not that spending less is always better.

Myth versus evidence

The social-video version

Once returns start diminishing, every extra dollar of ad spend is wasted.

What the evidence supports

Extra spend still pays while each additional result costs less than it earns; diminishing returns are not the same as negative returns.

Five-minute field assignment

Pick one channel and export weekly spend and orders (incremental if you have them) for the last six months. Sort the weeks into low, medium and high spend, and calculate the extra orders gained per extra $1,000 between the groups. Does the figure shrink as spend rises? Note anything else that differed in high-spend weeks, such as season or promotions.

Checkpoint

Two questions: one about the idea, one about a decision. Answer both correctly to master this law. Questions are drawn from a larger bank and graded on our server; a retry brings different questions.

Evidence ledger

Every factual claim in this lesson, where it comes from, and how far it has been checked.

Further reading

Which other law could explain this decision?

Editorial review: Ali Dezhami, November 8, 2026. Last substantive update: November 8, 2026. Our methodology