What Is ROAS? How to Measure Return on Ad Spend Correctly

ROAS (Return on Ad Spend) is the conversion value generated by advertising divided by the cost of that advertising. If a campaign spends 10,000 and produces 45,000 in tracked revenue, the ROAS is 4.5 — or 450 percent, depending on how your reporting expresses it. The formula is trivial. Almost every serious problem with ROAS lives somewhere else: in the value, in the attribution, or in the target.

The formula and its variants

ROAS and CPA are two views of the same transaction. CPA is appropriate when every conversion is worth roughly the same — a lead form, a consultation booking, a subscription at a fixed price. ROAS is appropriate when order values vary, which is almost always the case in e-commerce.

Calculating your break-even ROAS

Before you set any target, calculate the point below which advertising destroys value.

1. Determine gross margin per order: selling price minus cost of goods, payment processing and outbound shipping.
2. Divide that figure by the selling price to get the margin rate.
3. Divide 1 by the margin rate. That is your break-even ROAS.

The right-hand column is the one most advertisers skip. Breaking even on advertising is not a goal; it is a floor. Your target has to leave room for overheads, returns and the marketing that is not directly attributable.

Why most ROAS numbers are wrong

The conversion value is inaccurate

Common failures, in rough order of frequency:

• Value passed includes VAT or shipping. The platform optimizes toward gross order totals that you never keep.
• Returns are never deducted. Categories with high return rates — apparel, footwear — can show a healthy ROAS while contributing nothing.
• Coupon and discount amounts are not subtracted. The tag fires the pre-discount value.
• Every product is treated as equally profitable. A catalog with margins between 8 percent and 62 percent is optimized as if it were uniform.

Conversions are counted more than once

Duplicate tags, a GA4 purchase event firing alongside a Google Ads conversion tag on the same thank-you page, and a page that fires on refresh all inflate the numerator. The symptom is a ROAS that looks strong in the ad platform and cannot be reconciled with the payment provider.

Attribution is inconsistent

Each platform claims the conversions it can see. Meta reports on its own attribution setting, Google Ads on data-driven attribution within its window, and GA4 on its own model across all channels. Adding up platform-reported revenue across three channels routinely exceeds actual revenue.

The only durable fix is to nominate a single source of truth — usually the order data in your commerce platform or ERP — and treat each ad platform’s number as a directional signal for optimizing that platform, not as a business result.

From ROAS to POAS

For a catalog with uneven margins, ROAS actively misleads the bidding algorithm. If a smart bidding strategy is told that a 100 sale of a low-margin item is worth the same as a 100 sale of a high-margin item, it will happily spend more to win the low-margin one because it converts more easily.

Profit on Ad Spend fixes this by sending gross profit — not revenue — as the conversion value.

1. Extract cost of goods per SKU from your ERP or commerce platform.
2. Calculate gross profit per line item at checkout: price minus COGS minus processing and shipping cost.
3. Send that profit figure as the value parameter in the purchase event.
4. Keep the currency parameter consistent and document that the value now represents profit, so nobody misreads the reports later.
5. Rebuild target ROAS values against the new scale — a target of 4.0 on revenue might correspond to something near 1.3 on profit.

The migration is worth the effort when margin variance across the catalog is wide. If every product sits within a few points of the same margin, the added complexity buys very little.

Setting a target that works

Target ROAS bidding is a forecasting model. It needs three things before it performs:

1. Volume. A campaign producing a handful of conversions a month gives the model almost nothing to learn from. Consolidate campaigns rather than fragmenting them.
2. Value accuracy. Garbage values produce a confidently wrong bidding strategy.
3. A realistic target. Setting a target far above recent actual performance causes the system to suppress impressions rather than find magic efficiency. Start from the trailing 30-day actual and move in increments of 10 to 15 percent, then wait through the learning period before judging.

A practical rhythm: change one variable at a time, allow at least two conversion cycles between changes, and record every change with a date so that performance shifts can be traced to a cause rather than to intuition.

Where ROAS stops being the right metric

ROAS is a short-window efficiency measure. It is a poor guide in three situations:

• Long sales cycles. For a service with a 60-day consideration period, a 30-day window reports a fraction of the eventual value.
• Repeat-purchase businesses. If the second and third orders arrive through direct or email, first-order ROAS understates the channel that acquired the customer. Blend it with customer lifetime value by acquisition cohort.
• Growth phases. Pushing ROAS higher usually means spending less. A campaign at 8.0 ROAS and low volume may be worth less to the business than one at 3.5 ROAS with five times the profit in absolute terms. Efficiency and scale trade off against each other; the business decides which it needs.

For that last point, the useful discipline is to look at marginal ROAS: what did the last increment of spend return, rather than what did the average unit return. Average ROAS falls as you scale, and that is normal, not a failure.

A measurement checklist

• Conversion value excludes VAT, shipping and discounts, or is documented if it does not.
• Exactly one purchase event fires per transaction, verified in preview mode and in real orders.
• Enhanced conversions are enabled with hashed first-party data where legally permitted.
• Consent Mode is configured so that consent-denied traffic is modeled rather than lost.
• Returns and cancellations are reconciled against the payment provider at least monthly.
• One system is nominated as the source of truth for revenue.
• Break-even ROAS is documented and reviewed whenever pricing or supplier costs change.

Closing

ROAS is a useful number and a dangerous one. It is useful because it converts spend into a single comparable ratio. It is dangerous because that ratio looks equally authoritative whether the data behind it is clean or broken. Fix the value, fix the deduplication, agree on one source of truth, and only then start tuning targets.

Moon Workshop builds measurement infrastructure and campaign management as one engagement — GA4, server-side tagging, Merchant Center feeds and bidding strategy reviewed together — for e-commerce and service businesses in Türkiye and in international markets. If your ad platform and your accounting disagree about revenue, that reconciliation is where the work starts.