What is POAS?

Two campaigns each spend A$5,000. The first returns A$20,000 of revenue on products at a 25% margin, so A$5,000 of gross profit: a return on ad spend of 4.0 and a profit on ad spend of 1.0. The second returns A$12,500 on products at a 60% margin, so A$7,500 of gross profit: a return on ad spend of 2.5 and a profit on ad spend of 1.5. Ranked on revenue the first campaign wins by a wide margin. Ranked on what the business keeps, the second one does.

Return on ad spend divides revenue by cost. It is the default in every ad platform and it treats every dollar of revenue as interchangeable. In a catalogue where margins are even, that assumption is harmless. In one where they are not, it is the most expensive assumption in the account.

Profit on ad spend makes the obvious substitution: divide gross profit by cost instead. What changes is not the arithmetic but the ranking, and the ranking is what budgets are made from.

A campaign selling high-volume, low-margin goods can post an impressive revenue return and produce very little money. A campaign selling fewer, better-margin items can post a modest revenue return and produce more. Every account with a mixed catalogue has some version of this pair in it, and on revenue-based reporting the first one gets the budget.

Gross profit here means revenue less everything that scales with the order. At minimum:

Cost of goods, at the product level rather than a store-wide average. A blended margin across a catalogue with real variation defeats the purpose of the metric before it starts.

Shipping, both what it costs to fulfil and what was charged. Free shipping is a margin decision recorded as a delivery one.

Payment processing, which is small per order and not small across a year.

Expected returns, ideally by category. A blended return rate across a store selling both apparel and hard goods misprices both, usually badly, because the categories differ by more than most other assumptions in the calculation.

What stays out is anything fixed: rent, salaries, software, the advertising team. Those are covered by the surplus the metric measures rather than subtracted inside it.

The reason this metric is discussed more often than it is used is that computing it after the fact is easy and optimising toward it is not.

A spreadsheet can join the orders a campaign produced to the cost of the products in them and produce a perfectly good profit figure for last month. That is useful for allocation and useless for bidding, because the bidder never sees it.

To optimise toward profit, the conversion value sent back to the platform has to be profit rather than revenue. That means product cost has to reach the purchase event — usually through a server-side event that enriches the payload, or through a feed that carries margin data the platform can read. Most ecommerce platforms hold the cost data already; the work is in the pipe.

Once the pipe exists the change is small and the effect is not. A value-based bid strategy pursues whatever value it is given. Given revenue, it finds revenue. Given profit, it finds different orders, different products and frequently different audiences.

Three places, reliably.

Across the catalogue. Entry-level products with thin margins are the easiest things in any store to sell and the least worth selling. Shopping and catalogue campaigns gravitate to them because they convert, and a revenue-valued bidder is happy about it.

Across discount levels. A 20% discount code lowers margin by far more than 20% of the profit on a low-margin item, and revenue-based reporting barely registers it. Profit-based reporting registers little else.

Across channels. Channels differ in the product mix they sell, not only in their efficiency. A channel that sells the flagship at full price and a channel that sells clearance stock can report identical revenue returns and completely different contributions.

It is still attributed. Profit on ad spend inherits every attribution weakness of the metric it replaces. If the platform is over-claiming conversions, it is now over-claiming profitable ones. The blended and incremental views are still required.

It ignores fixed costs. A ratio of 1.0 means the advertising paid for itself and contributed nothing else. The useful break-even is higher, and where it sits depends on what share of overhead the advertising is expected to carry.

It says nothing about the customer's second order. A campaign that acquires low-margin first-time buyers who return at full price later can look poor on this metric and be the best thing in the account. Where repeat purchase matters, the profit figure has to be read against lifetime contribution rather than the first transaction.

It is only as good as the cost data. A stale cost of goods, a landed cost that ignores freight, or a returns assumption set once and never revisited will all produce a confident, precise and wrong number, and the bidder will pursue it.

The sequence that works in practice is unglamorous and short.

First, get product-level cost into the order record, if it is not already there. This is usually a data problem rather than a marketing one and it is worth escalating as such.

Second, compute the metric retrospectively in a report, by campaign and by product group, and look at the ranking against the revenue-based one. The gap between the two rankings is the size of the prize, and it is what justifies the engineering work for the next step.

Third, get profit into the conversion value and switch the bid strategy to optimise on it. Expect the account to spend differently within a fortnight.

Fourth, revisit the cost and returns assumptions on a schedule. A margin model that nobody owns will drift, and the drift is invisible because the number it produces still looks like a number.

Profit on ad spend is not a replacement for the rest of the panel. It is the allocation metric — which campaign, which product group, which audience deserves the next dollar.

The blended efficiency ratio remains the control, because it reconciles with the bank and cannot be double-counted. Contribution margin per order is what sets the ceiling. Payback period is what decides whether the business can afford to wait for the contribution it has bought.

The profit ratio's contribution to that panel is specific: it is the one that stops a business from scaling the campaigns that sell the most and keep the least.

• Get product-level margin into the conversion value before anything else. Without it there is no profit figure to divide, and every other step on this list is unavailable.
• Send profit as the conversion value to the bidder where the platform supports it. A value-based strategy optimising on margin buys different inventory from one optimising on revenue, and the difference is the whole point of the metric.
• Subtract what varies with the order, not only cost of goods. Shipping, payment fees and expected returns all scale with volume and all sit between revenue and what the business keeps.
• Rebuild the returns assumption by category rather than using one store-wide rate. Apparel and hard goods differ enough that a blended rate misprices both.
• Expect the campaign ranking to change and plan for the argument. A campaign that looks strong on revenue and weak on profit has usually been defended for a long time on the first number.