Why Marginal CPA Always Rises When Scaling

The Algorithm's Prioritisation

Google's bidding algorithms are not stupid. When you give them budget, they spend it on the most likely converters first. The easiest wins go first. The harder wins follow.

At low spend, you capture branded searchers, high-intent shoppers, and people actively comparing. These convert efficiently. As spend increases, you reach casual browsers, early-stage researchers, and tangentially interested audiences.

The progression is inevitable. There is no setting that says "keep spending only on the good ones." More budget means reaching further into less qualified audiences.

Calculating Marginal CPA

Marginal CPA requires comparing performance at different spend levels. Track conversions and spend week by week. Plot the relationship. The slope tells you what each additional pound delivers.

In this example, average CPA rose from £20 to £23, which seems manageable. But the marginal CPA of that extra £5,000 was £33: 65% higher than the original efficiency. That marginal pound is working much harder.

The Profitability Threshold

Every business has a maximum CPA that generates profit after accounting for product costs, fulfilment, returns, and overheads. When marginal CPA exceeds this threshold, additional spend destroys value.

The challenge is that average CPA can remain below threshold while marginal CPA exceeds it. You appear profitable overall while losing money on every new pound spent. The profitable core subsidises unprofitable expansion.

Campaign-Level Differences

Marginal CPA varies dramatically by campaign type. Branded search might maintain low marginal CPA until saturation because demand is finite and high-intent. Performance Max might show steep marginal increases as it reaches into Display and YouTube inventory.

Understanding which campaigns have favourable marginal curves allows smarter budget allocation. Scale the campaigns with flattest curves first. Accept steeper curves only when flat-curve campaigns are saturated.

Time-Based Variation

Marginal CPA is not static. It varies with seasonality, competition, and market conditions. During Black Friday, overall demand increases but so does competition, often steepening the marginal curve.

Regular recalculation is essential. The marginal economics that justified spend levels three months ago may no longer apply. What was profitable expansion might now be loss-making.

The Optimisation Trap

Optimising for average metrics while ignoring marginal economics is a common trap. You hit your average ROAS target. You hit your average CPA target. But your marginal performance is terrible.

This happens because the algorithm optimises for the target you give it across all spend, not just incremental spend. A 3:1 ROAS target might be achieved by mixing 8:1 and 1:1 performance. The 1:1 portion is destroying the value the 8:1 portion creates.