What metrics should be used to make acquisition cost projections based on the level of advertising spending?
While MTA is essential for the daily optimization of marketing campaigns , it is not designed to recommend budget allocations in the event of spending variations (budget increase or decrease).
MTA does not take into account diminishing returns , auction dynamics , or simply the reach still available on your campaigns.
This is where the concepts of incrementality , elasticity and MMM take on their full value.
Here are the main KPIs we will discuss and that you can use to define an optimized budget allocation :
Marginal Cost of Acquisition (mCPA)
→ How does my cost per purchase change if I spend a little more? If I spend a little less?
Incremental Cost of Acquisition (iCPA)
→ What is my actual cost per purchase (the one actually caused by advertising)?Supported by arigorous scientific approach.
Marginal Incremental Cost of Acquisition (MICPA)
→ What will my cost per incremental purchase be if I increase (or decrease) my spending.
Incrementality and elasticity are related but do not measure the same thing.
The common point between incrementality and elasticity
Both examine the effect of a change in media budget .
Both are used to decide how far to scale a campaign based on a target CPA.
In both cases, we seek to identify the point where spending more is no longer profitable .
The key difference
Elasticity → continuous, marginal: “If I add 1% of budget, how does the CPA or volume evolve?”
Incrementality → binary, causal: “Without the campaign, would I have had these conversions?”
Incrementality as a diagnosis of causality , and elasticity as a reaction law .
1. Understanding Incrementality
The question we ask ourselves: How many conversions are actually caused by the campaign ?
We compare the volume of sales made by two populations A and B (often organized by region).
Population A will receive special treatment, which will not be applied to population B.
Example :
Background: We are looking to see if YouTube spending is truly incremental.
Test: 50% of France will be targeted by the YouTube campaign ( population A ), and 50% of France will be excluded from the YouTube campaign ( population B )
Measurement: We isolate the causal part of sales (≠ organic, ≠ cannibalization).
It is a causal measure, which allows to measure an iCPA (incremental CPA).
iCPA = Average CPA / %_incrementality
We can also express the incrementality in lift vs control group:lift = (sales with treatment − sales without treatment) / sales without treatment = 200/800 = 25%.In this case :iCPA = Average CPA × (1 + lift) / lift = 10 × 1.25 / 0.25 = €50 .(Equivalent to the previous formula, just another way of writing it.)
2. Elasticity
The question we ask ourselves: How does CPA or sales volume react when I change my expenses ?
This is a sensitivity measure. We calculate the elasticity of conversions in relation to spending.
CPA elasticity = % variation of CPA when the budget is varied.
An elasticity of 0.5 means that sales increase, but half as fast as expenses .We gain weight less quickly than we spend.
Concretely, if we increase expenses by 10% , we only recover +5% of sales .
As long as the incremental value of a conversion is greater than the marginal cost, it is profitable, but we are approaching a zone of diminishing returns.
Example 2: Budget +10%, CPA −5% → elasticity =−0.5
Negative elasticity means: when you increase spending, the CPA decreases . Every additional euro makes the campaign more efficient.
With an elasticity of −0.5 , spending more improves performance : +10% budget reduces CPA by 5% — we are in a zone of increasing efficiency.
In other words, we are entering a zone of increasing returns : the cost per sale becomes more efficient with more budget.
We are probably in a phase where:
the countryside was not yet saturated,
or media algorithms find better opportunities with more volume.
Theelasticity of conversions with respect to expenditures allowsthe marginal CPA to be deduced via a simple relationship.
mCPA = Average CPA / Elasticity
The CPA elasticity mentioned here assumes a roughly linear relationship between expenditures and performance in the available data area .
In practice, the curve may flatten (diminishing returns), change slope (threshold/promotion effects) or react with a delay (seasonality, learning time).
Elasticity is a
simple and useful tool
, but its validity is
local
. For ambitious budget changes, choose a
nonlinear modeling
or controlled tests.
3. Operational recommendations
IF
THEN
Spending variation ≤ 15–20% and stable media environment
Use elasticity on attributed conversions
Variation in expenditure > 20% or multi-channel reallocation
Use elasticity on incremental conversions (holdout/geo-lift test) or Marketing Mix Modeling (MMM)
Strong seasonality/offline or major changes (algo, creative, offer)
MMM (with adstock + and saturation management)
In summary
Incrementality → “Does it work, and how much more does it create?”
Elasticity → “If I put a little more or a little less, how does it move?”
Link→ideally , we first measure the incrementality to be sure of the causality, then we calculate the elasticity to optimize the budget level.
However → incremental measurement tests are expensive and elasticity is often used independently.