Is Your Agency's ROAS Hiding Weak Business Performance?

A paid media agency reports a 4x return on ad spend and calls the quarter a success. Finance closes the same quarter and sees contribution flat or down. Both numbers can be correct at the same time, because ROAS divides attributed revenue by media spend and stops there. As Saras Analytics puts it, the ratio says nothing about whether revenue survives fulfillment, returns and operating costs. The ROAS vs profitability gap carries a real cost. When ROAS is the contractual yardstick, it quietly rewards revenue that may never cover cost of goods, shipping, discounts or the retainer itself. This article explains the mechanism, shows how to compute the break-even multiple your margins actually require, and lays out how to move an agency contract from platform ROAS to profit contribution without losing the optimization signal the agency still needs.

Most ROAS vs profitability arguments are framed as a metrics debate. They are really a contract debate. When a buyer names ROAS as the agency's yardstick, the buyer has picked a measure that answers how efficiently media became attributed revenue, while the decision on the table is whether this partner should keep this budget. Those are different questions with different denominators. GPI's position is that a strong ROAS is a claim like any other agency claim: check what it counted, what it excluded, and whether it reconciles to the ledger. Attribution shows association, not cause. Match the metric to the decision, count the fee, and test the increment before rewarding the multiple.

TL;DR: How ROAS Can Look Healthy While Profit Stalls

Short answer: yes, a healthy agency ROAS can coexist with weak business performance, and the reasons are structural rather than dishonest.

• ROAS divides attributed revenue by media spend. It never sees cost of goods, fulfillment, returns, payment fees or the agency retainer, so it can rise while contribution falls.
• A campaign only makes money above its break-even ROAS, roughly 1 divided by contribution margin. A 4x ROAS on a 20% margin product loses money before any fee is paid.
• Platform ROAS is an attribution estimate, not a ledger figure. Swydo's hypothetical of a 6:1 in-platform ROAS sitting on a 2.5:1 blended figure shows how wide that gap can be.
• Optimizing an agency to ROAS optimizes ad performance, not business performance, and the gap tends to widen as spend scales.
• The fix is contractual as much as analytical: agree margin tiers, reconcile to finance monthly, and put agency fees inside the efficiency calculation.
• Low ROAS is not automatically bad and rising ROAS is not automatically good. Judge both against margin, payback and incrementality.

ROAS vs Profitability: What Each Number Can and Cannot Tell You

The ROAS vs profitability distinction comes down to what each ratio puts in its numerator and denominator. Define the metrics once, then map each to the decision it can safely support.

What ROAS actually measures

Return on ad spend is attributed revenue divided by media spend, a formula that finance training sites such as CFI and Wall Street Prep present as a media-efficiency ratio. It is a good comparative tool: two campaigns, two audiences or two creatives can be ranked against each other because they share the same blind spots. It tells you which lever produced more attributed revenue per media dollar. It does not tell you whether that revenue was worth having.

What profitability metrics add: ROI, POAS and contribution margin

ROI subtracts all relevant costs from the return before dividing, so it answers whether the activity created value rather than how efficiently it produced revenue; one practitioner summary frames it as efficiency of spend versus overall profitability, and ecommerce tooling vendors draw the same line. Contribution margin is revenue minus variable costs. POAS, profit on ad spend, divides that contribution by media spend. The costs ROAS structurally omits are the ones that decide whether a sale helps or hurts: cost of goods, shipping and fulfillment, returns, payment processing, checkout discounts, and agency and creative fees, a list Saras Analytics groups under fulfillment, returns and operating costs.

Which question each metric is fit to answer

Formulas above are standard arithmetic; the ROAS definition follows CFI and the ROI contrast follows the LinkedIn summary cited above.

The metrics disagree by design, not by error. Different numerators, different denominators and different time windows produce different numbers from the same month of trading. The practical exercise is to open the current agency report, label each figure as efficiency or profitability, and note which budget or renewal decisions currently rest on an efficiency figure. GPI's guide to evaluating agency performance by profit covers how to structure that review.

When a High ROAS Campaign Is Actually Losing Money

To find out whether a reported ROAS is profitable, compute the break-even multiple your margin requires, add the costs the platform never sees, and compare. The steps below use only arithmetic; no figure here is an industry benchmark.

Break-even ROAS is set by contribution margin, not by benchmarks

Before fixed costs, break-even ROAS is approximately 1 divided by contribution margin. A 50% margin needs 2x, 25% needs 4x, 20% needs 5x. Whether a 4x is good therefore depends entirely on what you sell, which is why comparing an agency's number to a published benchmark answers the wrong question.

Take a hypothetical case. Assume $10,000 of media spend, a reported 4x ROAS, so $40,000 of attributed revenue, and a 22% contribution margin. Contribution is $8,800. Against $10,000 of media that is already a $1,200 loss, because break-even at 22% is about 4.5x. Add a 15% agency fee on spend, $1,500, and the month is $2,700 underwater while the dashboard shows a green 4x. These numbers are illustrative, not measured results, but the structure is what practitioners describe when they explain how a high ROAS can still lose money.

All table values are hypothetical and computed from the stated assumptions: fee column uses 1.15 / margin; the final column also treats 10% of attributed revenue as returned and recovers no margin on it, so it uses 1.15 / (0.9 x margin). Your own return handling and fee basis will change the figures.

Where the margin leaks after the sale: returns, shipping, discounts, fees

Platform ROAS records the order at checkout and never revisits it. Returns and chargebacks that arrive weeks later, promotional discounts applied at checkout, subsidized shipping and payment processing all sit between attributed revenue and contribution, which is the gap Saras Analytics describes when it separates revenue efficiency from profit. A product line with a high return rate can post a strong ROAS every month and still drain cash.

Scaling a profitable-looking ROAS can scale the loss

ROAS optimization can worsen the mix. Bidding algorithms and agencies alike will lean toward whatever produces the most attributed revenue per dollar, which often means high average order value SKUs or heavily discounted items. Revenue inflates, margin thins. Admetrics describes this as optimization that encourages ad performance rather than business performance. Double the budget behind a below-break-even ROAS and you double the loss while the ratio holds steady.

Signals that this is happening in your own account:

• Revenue is up quarter on quarter while cash and contribution are flat.
• The best-looking SKUs in the ad report are not the best-earning SKUs in the ledger.
• Discount depth or return rates rose in the same period the ROAS improved.

Compute your break-even per margin tier, add your actual fee, then place the agency's reported ROAS next to it. That comparison tells you whether the reported multiple is a win or a loss.

Why the Agency's ROAS and Your Finance Ledger Disagree

The agency's ROAS and your finance ledger disagree because they count different things over different windows. Understanding the four sources of divergence lets you request a reconciliation rather than argue about whose number is right.

Platform-attributed revenue is a model, not a receipt

Each ad platform claims conversions using its own attribution window and model. A single order can be claimed by search, social and an affiliate network, so summing platform-reported revenue routinely exceeds actual orders. Swydo's hypothetical illustration of a Performance Max ROAS of 6:1 masking a blended 2.5:1 across the business is a constructed example, but the direction of the gap is what overlapping attribution windows produce.

Siloed ROAS versus blended business efficiency

Blended efficiency, total ledger revenue divided by total marketing spend, sometimes called MER, has one virtue: it reconciles to finance. It cannot tell you which channel earned the revenue, but it is the number a CFO will sign off, which makes it the right anchor for the agency's platform figures.

The platform-versus-blended gap in the table follows the Swydo illustration cited above; the return and timing lags follow standard revenue recognition practice.

Privacy changes made the model more modeled

Consent requirements and browser and device tracking restrictions have removed part of the observed conversion path, and platforms fill the holes with modeled conversions. The ledger here contains no figure for how much of a given account is modeled, so treat this as a mechanism, not a quantity. LiveRamp's guidance on overcoming data gaps to prove ROAS exists because marketers now have to bridge gaps that did not exist a few years ago.

Media waste the ROAS report never shows

A ROAS report shows revenue per dollar; it never shows the dollars that bought nothing. A January 2024 opinion piece in The Drum cited research that 41% of overall ad spend goes to waste, a figure that traces to a sponsored Digiday report from April 2023. Read it as a historical, sponsored data point rather than a current benchmark. Its usefulness here is the reminder that a ratio can look fine while a large share of the denominator produced no return.

The practical routine: each month, place platform-attributed revenue, blended revenue and finance-recognized revenue net of returns side by side. Ask the agency to supply the first two columns for the last quarter and explain the gaps.

How Agency Fees and Contract Incentives Keep ROAS on Top

Agency fees are real acquisition costs, and the way most contracts treat them explains why ROAS stays on top even after everyone in the room agrees it is insufficient.

Fees sit outside the ROAS denominator

Retainers, percentage-of-spend fees and creative production are costs of acquiring the revenue in the numerator, yet none of them appear in the ROAS denominator. Add them and the picture shifts: a 15% fee on spend raises every break-even multiple in the earlier table, because the business is paying $1.15 for every dollar the platform thinks it paid. The omission is the same structural one Saras Analytics describes for fulfillment and operating costs, only this time the cost is the partner producing the report.

Percentage-of-spend compensation rewards scale, not margin

When the agency is paid on spend and judged on ROAS, growing spend into high-revenue, low-margin inventory is rational behavior. Revenue rises, ROAS holds, the fee grows. The client absorbs the margin loss. Admetrics' observation that ROAS optimization rewards ad performance rather than business performance becomes locked in once compensation is tied to the same efficiency ratio. No bad faith is required; the contract asked for this.

The table is a conceptual framework, not a survey of specific agencies' fee models; the incentive logic follows the Admetrics claim cited above.

What a ROAS-only contract asks the agency to ignore

Agencies report the metric the platform surfaces and the contract names. That is a design problem, not an honesty problem. A ROAS-only contract tells the agency that margin data is irrelevant, that returns are someone else's line, and that the fee does not count. The organizational cost lands in the leadership meeting: the CMO presents ROAS wins, the CFO presents flat contribution, and the argument is really about which denominator each side is using.

Before renewal, audit the agreement against four questions:

• Is the agency fee inside the efficiency calculation the agency is judged on?
• Are targets tiered by margin, or is there one blended ROAS number?
• Does the agency have access to margin and return data at SKU or category level?
• What happens to the fee if contribution falls while ROAS rises?

Draft the incentive changes now, so the conversation at renewal is about the mechanism rather than about whose number to trust.

Transitioning the Agency Contract From ROAS to POAS

To move an agency contract from ROAS to POAS, build margin tiers, feed profit values into bidding and reporting, restate targets around break-even and contribution, then reconcile monthly and adjust the fee logic. POAS is gross profit after variable costs divided by ad spend, a definition consistent with how measurement vendors such as JENTIS describe it. Moving the numerator from revenue to contribution means the target itself carries margin, so the agency can optimize toward profit on ad spend without a finance analyst translating every week.

Step 1: Build margin tiers the agency can act on

1. Ask finance for contribution margin by SKU or product group, including cost of goods, shipping and payment costs.
2. Add return rates by category and the average discount depth per promotion type.
3. Collapse the result into a small number of tiers, for example low, mid and high margin, so the agency can map campaigns and product feeds to them.
4. Share the tiers under confidentiality; the agency needs ranges and rules, not the full cost ledger.

Step 2: Feed profit values into bidding and reporting

1. Pass contribution rather than revenue as the conversion value where the platform supports value-based bidding, or apply a margin multiplier per feed label.
2. Rebuild the agency dashboard so every campaign row shows spend, attributed revenue, estimated contribution and the fee allocated to it.
3. Confirm the platform POAS figure and the finance contribution figure are labeled differently, so nobody reads the first as the second.

Step 3: Rewrite targets around break-even and contribution

1. Replace the single blended ROAS goal with per-tier break-even ROAS floors, or with one POAS target above 1 after fees.
2. Illustrative restatement, hypothetical: a flat 4x ROAS goal becomes a 5x floor for 20% margin lines and a 2.5x floor for 45% margin lines, where break-even is about 2.2x and the extra margin covers fees.
3. Write the fee basis into the target sheet so the agency and finance compute the same floor.

Step 4: Reconcile monthly and adjust the fee logic

1. Run platform POAS weekly for optimization decisions.
2. Run ledger-reconciled contribution monthly for evaluation, net of returns that have landed.
3. Tie any bonus or spend-growth trigger to the monthly reconciled number, not the weekly platform number.
4. Review the tiers quarterly as costs and promotions change.

GPI's buyer guide on choosing a performance creative agency covers the selection-stage questions that make this handover easier later.

What POAS still cannot see

POAS improves the numerator but inherits the attribution problem described in the reconciliation section: the platform still decides which orders count. It also ignores new-versus-returning customer mix and fixed overhead. A campaign can post a POAS above 1 by harvesting repeat buyers who would have purchased anyway. POAS therefore complements blended efficiency and incrementality testing rather than replacing them.

Measuring True Business Return Beyond the Platform

Settling the ROAS vs profitability question takes a layered measurement stack, with a named owner and cadence per layer. The layers below build upward from the platform to the board.

Blended efficiency and reconciled contribution

The board-level view is total marketing spend, fees included, against ledger revenue and contribution. This is the layer that resolves the CMO and CFO argument because both start from the same ledger. It also protects against the kind of gap Swydo illustrates between a platform multiple and the blended business figure: if platform ROAS climbs while the blended ratio stays flat, attribution is claiming revenue the business did not gain.

CAC payback and first-order versus lifetime economics

CAC payback asks how many months of contribution from a new customer repay acquisition cost plus fees. A low first-order ROAS can be sound when payback is short and repeat rates are high; a high first-order ROAS can be poor when most buyers never return. Separating new-customer efficiency from remarketing efficiency matters here, because blended ROAS can be propped up by returning buyers the agency did not create. Practitioner comparisons of ROAS and ROI make the same point about efficiency versus full-cost return without settling which window to use, which is a decision your own repeat-purchase data has to inform.

Incrementality checks before you trust the trend

Attributed revenue is not proof the ads caused it. Holdout groups, geo splits or scheduled spend pauses estimate how much revenue would have occurred anyway. A quarterly test is enough for most mid-market buyers to calibrate how much of the platform's claim to believe.

The stack is a conceptual operating model; no benchmark values for payback or MER are given because none in the ledger are verified.

Assign an owner to each layer and identify the one currently missing. Agency-produced ROAS reports seldom include the ledger or causal layer, so check those first.

When a Low ROAS Is Fine and When a Rising ROAS Should Worry You

The same ROAS movement can mean opposite things depending on margin, mix and scale. Three patterns cover most cases.

Low ROAS, healthy business: high-margin and subscription models

A subscription, high-repeat or high-margin product can run below common ROAS expectations and still be profitable, because payback arrives over subsequent orders rather than the first one. Judging such a business on first-order ROAS punishes exactly the acquisition that builds it. The right tests are contribution margin and CAC payback, not the multiple.

Falling ROAS while scaling: expected, up to a point

Pushing spend into new audiences typically lowers marginal ROAS through auction pressure and audience saturation. Guides that treat a ROAS decline as an early warning are right that it deserves attention, but the question to ask is whether marginal contribution stays above zero after fees, not whether the average ratio fell. Brief the agency to report marginal efficiency, the return on the last increment of spend, whenever budgets move, since average ROAS hides the shape of the curve.

Rising ROAS with shrinking contribution: the warning sign

The inverse pattern is the dangerous one. When ROAS improves while contribution falls, the mix has usually shifted toward discounts, returning customers or low-margin volume, the ad-performance-over-business-performance drift Admetrics describes. Analytics vendors writing about declining platform ROAS versus true profitability make the related point that the platform figure and the profit figure can move independently.

The table is a diagnostic framework, not a set of client cases; the mix-shift mechanism follows the Admetrics claim cited above.

Classify your last two quarters into one of the three rows and request the matching metric from the agency before the next review.

How GPI Reads Agency ROAS Claims

A reported ROAS is a claim. The buyer's job is to check what it measured, which costs it excluded and how it reconciles to finance. That is the same lens GPI applies to any agency claim in its methodology: evidence, method and limitations, in that order.

Ask for the denominator and the reconciliation, not the multiple

The decision in front of a buyer is concrete: keep or change this agency and its budget. Choose the metric that answers that decision rather than the one the platform surfaces by default. In practice this means asking two questions before discussing whether a multiple is good. What is in the denominator, and does the fee sit inside it? How does the platform revenue reconcile to the ledger over the same period? An agency that cannot answer either is reporting ad performance, not business performance, whatever the number says.

Evidence, methodology and limitations as the evaluation lens

Attributed revenue shows association. It does not show that the agency's creative and media choices caused the contribution, which is why a gap like the platform-versus-blended contrast Swydo illustrates should prompt a test, not a celebration. For a shortlist or renewal, ask each agency about:

• Willingness to work from margin data and report by tier
• Fee-inclusive efficiency reporting
• Monthly reconciliation of platform revenue to blended and ledger revenue
• Appetite for incrementality tests and how results would change spend
• How the agency's own case studies define return, and whether those definitions include cost

Browse how GPI documents agency evidence in profiles such as The Goat Agency and ACE Agency before your next review. These profiles organize documented evidence about agencies; they do not present profit-based case results, and GPI has not run campaigns with any listed firm. The remaining decision is whether to retain ROAS as the contract yardstick or rewrite the performance clause around break-even by tier, reconciled efficiency and contribution net of fees before renewal.

FAQ: ROAS vs Profitability for Agency Buyers

How do we calculate break-even ROAS when margins differ widely across SKUs?

Group SKUs into a few margin tiers and compute 1 divided by the tier's contribution margin, then add the fee as shown in the break-even table. Run separate campaigns or feed labels per tier so each carries its own floor. A single blended break-even hides the losers.

Should the agency's fee be included in POAS or reported separately?

Report both. Platform POAS without the fee is the agency's optimization signal; POAS after fee is the evaluation number and the one any bonus or spend trigger should reference. Showing them side by side prevents the fee from disappearing.

What data does an agency need from finance to optimize toward profit, and how do we share it safely?

Margin ranges by tier, return rates by category and standard shipping and payment cost assumptions. Share tiers and multipliers rather than unit costs, under the confidentiality terms of the agreement, and refresh quarterly.

How long should we run both ROAS and POAS reporting in parallel before changing targets?

Long enough to see at least two monthly reconciliations line up with finance and to confirm the tiers are mapped correctly. Once the agency and finance agree on the reconciled contribution figure, switch the contractual target.

If platform ROAS is modeled, which number should the performance clause reference?

Reference reconciled contribution or blended efficiency including fees, with platform POAS as the operating metric the agency manages week to week. The clause should name the ledger-based number because it is the one finance can verify.

How do we account for returns that arrive weeks after the attributed sale?

Apply a category return-rate assumption to platform figures for weekly decisions, then true it up in the monthly reconciliation once actual returns land. Recompute the break-even floor when observed return rates drift from the assumption.

Does moving to profit-based targets reduce spend volume, and how should we set expectations with the agency?

Often spend falls on low-margin lines and rises on high-margin ones; total volume may drop initially. Set expectations by pairing the profit target with a marginal contribution floor and a minimum spend commitment, so the agency is rewarded for profitable scale rather than for shrinking to protect a ratio.