PPC budget: why double the money never doubles the orders

Campaigns are running, targets are holding, and someone in the meeting says: “Let’s add more.” It is the moment a good account most often gets broken — not because adding money is wrong, but because the decision is made on a number that cannot carry it. That number is average cost per conversion.

This article explains which number you need instead, why it is always higher, and how to work it out. It is the most useful arithmetic in all of PPC and it fits in three lines.

Why the return on added money falls

The first units of spend go to the cheapest and most willing customers: people searching for your brand, a specific product, or returning to an abandoned cart. Every further unit goes to a wider audience, into dearer auctions and onto less certain searches.

Two things happen at once, and it is useful to separate them, because both can be checked in the account:

• The click gets dearer. Cheap auctions are exhausted, dearer ones remain. Bids have to rise for the ad to appear at all.
• The conversion rate falls. Queries move further from the purchase. “Order pizza delivery in Camden” becomes “what is the best pizza”.

Put those together and you get the classic power-law advertising response. Written out, where K is conversions and S is spend:

K(S) = A · S^e, where the exponent e lies between zero and one.

The exponent e expresses how fast the market saturates. At e = 1 double the budget would give double the conversions; that does not exist in advertising. In practice, search advertising most often lands somewhere between 0.6 and 0.8. The difference is dramatic:

Marginal CPA: the number that decides

Now the useful part. Differentiate that power function and one very convenient thing falls out:

Marginal CPA = average CPA ÷ e. Marginal ROAS = average ROAS × e.

Marginal CPA is the cost of the conversion brought by the last unit of budget. At e = 0.7 it is therefore 43 % higher than the average CPA in the account. At e = 0.6 it is 67 % higher.

An example. A shop spends €4,000 a month and its average cost per order from ads is €20. Gross margin on an order is €24, so the account looks profitable — and it is. But marginal CPA at e = 0.7 is €28.60. Which means the last units of spend are buying orders for more than the margin. The account as a whole earns, the added budget loses, and nobody will see it in an average report.

Where the budget that maximises profit sits

The same formula answers the question of how much is “right”. Profit peaks where the gross margin from a marginal conversion equals one unit spent. In practice:

1. Work out gross margin on one conversion from ads. For a shop, margin on the average order; for an enquiry, margin on the job multiplied by the share of enquiries that close.
2. Compare it with marginal CPA, that is average CPA divided by the exponent.
3. While marginal CPA sits below the margin, adding pays. Once it exceeds it, every further unit reduces profit — even though revenue keeps growing.

The gap between the revenue-maximising budget and the profit-maximising one is large. Revenue grows until the advertising stops earning anything at all; profit stops growing much sooner. Companies that measure their account on revenue therefore almost always spend more than they should — and do not notice, because revenue really is growing.

You can have both calculated in our budget calculator: it shows marginal CPA, the profit optimum, and the budget at which profit from advertising reaches zero.

Checklist: twelve points before raising a budget

The model tells you whether it is worth adding. This checks that the business can take it. We use it ourselves, on every increase.

Data and business

1. Measurement matches real orders. Compare conversions in the ad platform with orders in the shop over the same period. Duplicates or counted cancellations can turn an average campaign into a star.
2. Stock and delivery can take it. More advertising means more orders and someone has to ship them. Advertising stock you do not have pays for angry customers.
3. You know margin by category. One cost-of-sale target for a whole shop works until you start scaling. Then money flows to where it sells easiest, not where it earns most.

Do the campaigns have room to grow

4. Impression share and the reason for losing it. What matters is why you lost impressions — lost to budget reads very differently from lost to rank.
5. The campaign actually hits its daily budget. If it does not spend it, something else is holding it back: too strict a target, narrow targeting or a weak feed. Added money changes nothing.
6. Brand is separated from non-brand. Otherwise you are raising budget against a distorted average. See brand vs. non-brand.

Is now the right moment

7. Learning phase. After a larger change of budget or target, automated strategies look for a new equilibrium. Do not give the algorithm two signals at once.
8. Seasonality. An increase a week before Black Friday and one in January are two different decisions. See seasonality in PPC.

Where people land

9. Landing page and checkout. Go through the purchase as a customer, on a phone. Every hole gets bigger with the increase.
10. A healthy product feed. Disapproved products mean part of the range never appears and the budget ends up on a handful of items.

How to raise so it can be evaluated

11. Expected result and review date written down first. One sentence, before you make the change. Without it, a week later there is always a reason why it actually went fine. We log every intervention including who approved it and when it gets reviewed — which is why we built ClientDeck.
12. In steps. A large jump disrupts the bidding strategy and makes it impossible to see where the return breaks. Add gradually and evaluate after each step. While the step meets the target, continue. The moment it stops, you know where the ceiling is.

How to find your own exponent

The value of e differs for every account and the model cannot guess it. But it can be measured, and not with any great difficulty:

1. Pick a campaign that regularly spends its budget and loses impressions to budget.
2. Raise the budget by 30–50 % and let it run at least two weeks after the learning phase settles.
3. Calculate by what percentage conversions rose and by what percentage the budget did. The exponent is roughly the ratio of those relative changes: e ≈ ln(K₂/K₁) ÷ ln(S₂/S₁).

If you get 0.4, the market is exhausted and the money belongs elsewhere. If you get 0.9, you have plenty of room. For bigger decisions a geo experiment is more reliable, because an ordinary budget test cannot tell new orders from orders that would have arrived anyway.

What to take away

A budget is not a growth lever, it is an amplifier. It will amplify a good setup as readily as a bad one. And above all: the decision to raise it is never governed by the account average. It is governed by marginal CPA compared with your margin — and that is always tens of per cent higher than the report suggests.