The 11x ROAS account that lost money on every order

An 11x return on ad spend (ROAS) sounds like a win. It can represent an outstanding ecommerce business. It can also represent a business quietly destroying cash.

The difference is context. It’s probably the most overused phrase in marketing, other than “it depends.” It’s also the one people ignore most often.

When 11x ROAS isn’t what it seems

We took over an ecommerce account that was reporting an 11x blended ROAS. The previous agency had it on a slide. The founder had quoted it to investors. Somewhere in the building, a finance director was staring at a cash flow forecast and quietly losing his mind because he was the only person in the company who knew both numbers.

The account was losing money on every order. Not the bad orders. The average order. After the costs ROAS has never heard of, the contribution per order was negative, and it had been negative for some time.

The figures below are a composite. The account is real, the pattern is real, and the numbers have been adjusted enough that no client is identifiable.

Every step of the math is shown so you can run it against your own account. I’d recommend you do. Some of you are about to have a bad afternoon.

Dig deeper: Why high-ROAS campaigns don’t always deserve more budget

What 11x actually claims

An 11x ROAS makes exactly one claim: ad spend was about 9% of reported conversion value. That’s it. That’s the entire confession. Everything else people read into it — the profitability, the health of the business — is inference.

The trouble is what lives inside “reported conversion value” and what never makes it into “ad spend.”

Here’s the stack on a £100 apparel order. Conversion value includes VAT (UK-based) and is booked before returns, which is how nearly every Shopify and GA4 setup reports it. Not because anyone decided that. Because nobody decided anything.

• Reported conversion value: £100.00
• Returns at 28%, booked after the conversion and invisible to the platform: £72.00 kept
• VAT removed: £60.00 net revenue
• COGS at 63% of net, a markdown-heavy trading period: £22.20 left
• Fulfillment, outbound shipping subsidy, return postage and handling: £11.20 left
• Payment and platform fees: £8.70 left
• Ad cost at 11x ROAS: -£0.39

£11 back for every pound in, yet the company loses £0.39 on every £100 order. The dashboard and the P&L describe the same orders. Only one of them is under oath.

The 28% return rate is unremarkable for apparel. A 63% COGS rate is what promo-heavy trading looks like, which is exactly when ROAS peaks because discounted stock converts beautifully. The account looked its best at the precise moment it was performing its worst, and nobody noticed because the metric they were watching was incapable of showing it.

The blend was doing the rest

That per-order math would be damning enough if the 11x were honest. It was blended.

Brand campaigns ran at roughly 18x and took most of the budget. Nonbrand sat around 3x. The headline figure was a weighted average of demand the brand already owned and a modest amount of actual work.

A decent share of those brand orders would have arrived through organic or direct anyway. People who type your brand name into Google have, broadly speaking, already decided to buy from you.

The true incremental return was well below 11x before the contribution stack even got involved. The account was taking credit for sales it didn’t cause, at a margin it didn’t have. As business models go, it’s a strong one. It just wasn’t the client’s.

None of this is exotic. It’s the default state of an ecommerce account that has spent a few years being optimized to platform-reported revenue. The reporting isn’t lying. It’s answering a very narrow question while everyone in the room treats it as the answer to a broad one.

Dig deeper: Why separating brand and non-brand campaigns improves ROAS

The part nobody warns you about

The textbook correction is to rebuild conversion values around contribution and bid to profit instead of revenue. We did this across our client base years ago. It works. POAS over ROAS is close to consensus now among people who think about ecommerce PPC properly.

Here’s what we learned from living with profit bidding for longer than most, and it’s the part I haven’t seen written anywhere: lean too hard into efficiency and you strangle the business from a different direction entirely.

A tight efficiency target tells Smart Bidding to buy only the cheapest, most certain conversions. It obliges. Volume drops. For a service business, that trade might be acceptable. For a retailer holding stock, it’s a serious problem because stock is a depreciating asset that happens to sit in a warehouse rather than anywhere a marketer might look at it.

A fashion brand pays for its season months before it sells it. Every week a unit sits unsold, its recovery value falls and the eventual markdown gets deeper. Whatever survives the season gets cleared below cost or carried at a cost. And next season’s buy is funded by this season’s cash. An account tuned for maximum efficiency sells fewer units and quietly converts working capital into boxes.

Consider a SKU with eight weeks left in season: 1,000 units at an £18 unit cost and £45 RRP. At a tight efficiency target, it sells 350 units at a strong margin, leaving 650 to be cleared at 70% off after the season, mostly below cost.

Relax the target to 4x, and it sells 850 units at a lower per-unit margin, leaving just 150 to clear. The second path produces a worse ROAS but more total contribution, while getting cash back into the business in time to fund next season. The efficient setting was the expensive one.

So the account in the title managed to lose money twice. On every order it made, through the contribution stack. And on every order it didn’t make, because it was too expensive to pursue. The first loss shows up in the P&L. The second doesn’t, because there’s no report for the sales that never happened.

Dig deeper: Your paid media ROAS isn’t 5x. Or 2x.

How to optimize Google Ads for profit and cash flow

The solution comes down to four changes in how you set targets and manage the account.

Rebuild conversion value as contribution

Net of VAT, expected returns by category, COGS, and fulfillment. Send that to Google as the value it bids against. Until the algorithm can see margin, every target you set is being negotiated in a fictional currency.

Separate brand before judging anything

Brand harvest at 18x tells you nothing except that people like the brand. Grade nonbrand on its own numbers, ideally on new-customer contribution.

Determine one job per SKU

An SKU can be run for profit, run for volume, or run to get cash back before the season closes. It can’t do two of those at once. The bidding strategies actively conflict. We reassign each SKU’s job weekly against the P&L. It is deeply unglamorous, and it’s the single change that got our clients’ ad accounts and finance teams speaking to each other again.

Decide the cash question out loud

For a stock-holding retailer, target setting is a weekly commercial call on where each SKU sits between margin, sell-through, and cash recovery. Sometimes the less profitable sale is the smarter move. That decision belongs to the business, not to a blended tROAS somebody set in January.

Dig deeper: Building high-ROAS ecommerce search campaigns in Google Shopping and Amazon Ads

When lower ROAS is the better result

When we restructured the account, blended ROAS fell. We separated brand, rebuilt values around contribution, and set volume targets against sell-through. The reporting got uglier, but the business got healthier. Contribution per order turned positive and cash began arriving during the season rather than at the clearance sale.

If your account is posting a number that would embarrass a hedge fund, find out what the number is actually measuring before it goes on a slide. Because an 11x on the wrong catalog is mostly a list of everything the account declined to sell.