POAS: Profit on Ad Spend, and When It Beats ROAS

Two campaigns close the month at a ROAS of 4.0. The first sells a product carrying a 55 percent gross margin. The second sells one at 18 percent. On the dashboard they are the same campaign. In the bank they are not close. At $10,000 of spend the first hands back roughly $22,000 of gross profit and the second hands back $7,200, which is less than the media invoice that produced it. ROAS cannot separate them, because ROAS counts revenue, and revenue does not pay suppliers.

POAS, profit on ad spend, closes that gap by replacing the numerator. Same shape of ratio, different top line, and a very different set of decisions falling out of the other end.

The formula, and the input that decides everything

POAS = (revenue - COGS - variable costs) / ad spend

Work an example. Ads produce $40,000 of revenue on $10,000 of spend, so ROAS reads 4.0. Cost of goods on those orders is $22,000. Payment processing takes $1,200 and pick, pack and shipping takes $2,800. Gross profit is $14,000, so POAS is 1.4. Every dollar of media returned $1.40 of gross profit, leaving 40 cents of contribution towards rent, salaries, software and everything else that exists whether or not the campaign runs. Break-even sits at a POAS of 1.0, and 4.0 was never the relevant number.

Which costs belong inside that bracket is the whole argument. Anything that scales with the order goes in: goods, processing fees, fulfilment, packaging, the realistic cost of returns. Anything that would exist at zero orders stays out. Overhead pushed into COGS makes every campaign look unprofitable and eventually gets a working channel switched off. Draw the line once and keep it, and if the boundary is unclear, settle it in the contribution margin calculator before you start reporting POAS to anyone.

How POAS and ROAS divide the work

ROAS is not broken. It is a revenue efficiency measure and it is honest about being one. If every product in the catalogue carries roughly the same margin, ROAS and POAS move together and one target serves both. The moment margin varies across the range, a single ROAS target starts quietly rewarding the wrong SKUs, because the cheap high volume line clears the target far more easily than the item that funds the business.

The usual patch is a per product break-even ROAS, which is just one divided by gross margin: a 55 percent margin needs 1.82 to wash its face, an 18 percent margin needs 5.56. That works, and the break-even ROAS calculator will produce the thresholds. It also means maintaining a different target for every group of products and re-checking them each time a supplier price moves. POAS states the same thing once. One threshold, 1.0, applies to the entire account, and the margin differences are already inside the number rather than sitting in a spreadsheet beside it.

Keep both on the report. ROAS is the language the platforms, the agency and most benchmarks speak, and you still need it to argue with anyone outside the business. Run your existing figures through the ROAS calculator and then put POAS next to it in the same table. Where the two disagree is where the money is.

What counts as a good POAS

Treat those bands as directional and nothing more. They come from accounts with different overhead ratios, different return rates and different definitions of variable cost, and none of them knows what your fixed costs are. A POAS of 1.3 in a business with almost no overhead can be comfortable. The same 1.3 with a warehouse and fourteen salaries behind it is a loss wearing a green cell.

Four places POAS breaks

The first is attribution. POAS inherits whatever revenue the reporting handed it, so platform-claimed revenue with a seven day click window produces platform-flattered profit. Nothing about switching to profit fixes over-counting, and if two channels both claim the same order, both will now claim the margin on it as well. A blended sanity check against total gross profit, the kind the MER calculator produces, is worth running monthly for that reason alone.

The second is returns. A 25 percent return rate on apparel does not simply remove the revenue. It leaves you with outbound shipping, return shipping, and an item that may be sold at a discount or written off. Most product feeds carry COGS and stop there, so POAS built straight off the feed overstates profit for exactly the categories with the worst return behaviour. Subtract a category level return allowance before you trust a single campaign level figure.

The third is discounting. Margin varies per order, not per product, the moment codes are in play. A 20 percent code on a 40 percent margin item halves the contribution, and the campaign that drove the most discount redemptions will look strongest in ROAS and weakest in POAS. That is the correct answer, but only if the discount is deducted in the calculation rather than sitting in a separate promotions report.

The fourth is time. POAS on the first order says nothing about the second, and a subscription or replenishment business that judges acquisition on a single order will underspend against a competitor who reads the whole customer. Where repeat purchase is real, pair the figure with the number from your LTV calculator and decide deliberately how much first order loss the cash position can carry.

Feeding profit back into the bidding

Measuring profit and bidding on revenue is a half-finished job. Both Google Ads and Meta accept a custom conversion value, so you can send gross profit instead of order value and let target ROAS bidding optimise on margin without knowing that is what it is doing. A target of 1.4 on profit values behaves as a POAS target, and the algorithm stops chasing the cheap high volume lines that were dragging the blended figure along.

Two cautions before you switch. Conversion values shrink by whatever your margin is, so historical targets are meaningless afterwards and every automated rule referencing them needs rewriting on the same day. And smart bidding needs volume: if a campaign generates fewer than about thirty conversions a month, changing the value definition mostly adds noise to a model that was already short of data. Fix the tracking, run it in parallel for a full purchase cycle, then move the targets.

Do this before your next budget review

Export the last 90 days of ad-driven orders with COGS attached, subtract processing, fulfilment, discounts and a return allowance, then divide by spend at campaign level and sort ascending. In most accounts the bottom of that list is one or two campaigns with a respectable ROAS and a POAS under 1.0, and they have been funded quietly for months because nobody looked past the revenue column. That single sort usually pays for the afternoon it takes.