ROAS vs ROI: Reconcile Media Revenue, Total Costs and Contribution

A campaign can report strong return on ad spend and still leave very little money for the business. ROAS compares attributed revenue, or another conversion value, with media spend. ROI compares a defined return with a defined investment. Those calculations answer different questions, and neither becomes reliable until the revenue, costs and time window agree.

The useful question in a ROAS vs ROI comparison is therefore: what happens between the revenue credited to advertising and the money remaining after the relevant costs? This guide builds that reconciliation step by step. It uses one entirely fictional Canadian retailer, explicit formulas and practical checks for refunds, margins, missing information and delayed sales. All monetary examples are hypothetical Canadian dollars, excluding sales tax. None represents a Canada Create client, a performance benchmark or a forecast.

What is the difference between ROAS and ROI?

Revenue ROAS is attributed revenue divided by ad spend. A result expressed as a multiple tells you how many revenue dollars were attributed for each dollar spent on media. It does not subtract the cost of supplying the product, delivering the service or running the campaign. A positive revenue figure will produce positive ROAS whenever media spend is positive, even if the business loses money.

ROI needs a more precise label because people use different cost boundaries. In this guide, total in-scope cost ROI means contribution after marketing divided by the combined variable fulfilment and marketing costs included in the reconciliation. This is an analytical measure for a defined campaign cohort. It is not company-wide net profit, a cash return, an annualized investment return or proof of advertising’s causal effect.

Google’s ROI explanation relates return to costs and illustrates including production and advertising costs. Its simplified example is not a complete accounting policy. The formula below makes our chosen boundary explicit so a reader can reproduce the result and see what remains outside it.

The final row is a separate, explicitly named diagnostic. It must not be silently substituted for total in-scope cost ROI. Identical revenue and costs can produce different percentages solely because the denominator changed. A responsible comparison prints the equation beside the result instead of expecting the word “ROI” to settle the definition.

Check what the reported conversion value actually represents

A dashboard labelled ROAS may use purchase revenue, estimated lead value, contribution value or a mixture of conversion values. Before describing the result as revenue return, identify the conversion actions included and the meaning of their values. A newsletter registration with an assigned value is not a completed sale, and a quote request is not collected revenue.

Google describes Target ROAS in terms of conversion value relative to advertising cost. That distinction matters: the calculation inherits the value definition supplied to it. If values represent estimated profit, call the result a profit-value-to-media-cost ratio. Do not insert it into a revenue reconciliation and deduct the same fulfilment costs again.

For a purchase-based comparison, write a short revenue policy before entering numbers. State whether order values already reflect discounts; whether delivery charges are included; whether refunds have already been deducted; and whether sales tax is excluded. Use one consistent treatment across the revenue export, order records and cost schedule. Document unresolved differences rather than adjusting a total until two systems happen to match.

Keep estimated lead value separate from realized sales

A service business may need an expected-value model while sales are still developing. Keep it in a planning column, with its qualification and close-rate assumptions visible. When actual sales arrive, replace or reconcile the estimate; do not add actual sales on top of the expected value of the same opportunities. A qualified enquiry, signed contract, delivered service and paid invoice are different milestones.

For this guide’s fictional retailer, gross attributed order revenue means accepted order value after any checkout discount and before refunds, excluding sales tax. No lead values, projected repeat purchases or unfulfilled pipeline estimates enter the example. This narrow definition allows the arithmetic to be checked without inventing a lifetime-value assumption.

Use one cohort, declared observation cut-offs and one cost boundary

“April performance” can mean April ad interactions, April orders, April invoice dates or April cash receipts. Those populations overlap, but they are not interchangeable. Pairing the cost of acquiring one group with the revenue from another creates a calculation that can look precise while answering no clear business question.

Google’s conversion reporting documentation distinguishes primary conversion columns based on ad-click timing from columns based on conversion timing. Its conversion-window documentation separately describes how long after an interaction a conversion can be counted. Record both the reporting date basis and the eligibility window; choosing identical calendar dates in two exports does not establish a match.

The hypothetical example follows advertising interactions from 1–30 April 2026. It includes associated orders observed through 31 May and refunds and cost adjustments observed through 30 June. These are illustrative cut-offs in the America/Toronto time zone, not recommended settings. Revenue and fulfilment costs belong to that interaction cohort, even when the order or refund occurs later. Marketing costs are assigned to the same cohort under a documented allocation rule.

A cohort is simply the group being followed. The workbook should give it a unique name and preserve the same name on every supporting schedule. Record the revenue eligibility rule, observation cut-off, extraction date, currency and reporting time zone. Distinguish costs incurred for that cohort from costs paid during its calendar month. A June payment for April creative does not automatically belong to June acquisition performance.

Keep provisional periods visibly provisional

If the observation cut-off has not passed, label the result provisional. Do not rank an immature cohort against one that has already accumulated later sales and refunds without explaining that difference. Retain earlier snapshots so the reader can see whether a change came from additional outcomes, a revised cost allocation or a correction to the underlying data.

A mature cut-off is also not a claim of perfect completeness. Late disputes, missing orders and revised invoices may remain. Establish a rule for reopening a cohort and recording the adjustment. A useful report says what was known at the extraction date and who owns outstanding items. It does not conceal uncertainty behind a rounded percentage.

Separate media spend from the costs of earning the revenue

Start with media spend as its own line. Keep campaign management, creative production and measurement costs outside that line so revenue ROAS remains recognizable. Then add those amounts to the broader marketing-cost total used by the ROI reconciliation. Renaming an agency invoice “ad spend” makes comparisons harder if another report excludes it.

The second cost group is variable fulfilment: costs that change with the goods or services supplied. Depending on the business, that may include inventory, packaging, payment processing, delivery subsidies, sales commissions or incremental delivery labour. BDC’s variable-cost explanation distinguishes costs that move with output from fixed and mixed costs. The classification still needs to follow the business’s actual arrangements.

Build a cost register with an identifier, description, category, amount, source record, service period and allocation basis. Mark whether each amount is directly attributable, allocated or excluded. A cost can be relevant even when it is not shown inside the advertising platform. Equally, a general corporate expense should not be assigned to a campaign simply because a spreadsheet has an empty row.

Allocate management and creative consistently

For the fictional retailer, the included non-media marketing costs are CAD 1,500 for management, CAD 1,000 for creative and CAD 500 for measurement. These are hypothetical allocations to the April cohort. They are not Canada Create prices, market rates or suggestions about what a business should pay.

For shared work, select a defensible basis before looking at which campaign benefits from it. Recorded hours may suit management work. A defined usage period may suit reusable creative. A direct assignment may suit an asset made for one campaign. Explain the method and reconcile allocations across all beneficiaries so the same invoice is not charged in full to several campaigns.

An allocation and an incremental cost are different ideas. Existing salaried staff may consume time without changing immediate payroll. A new contractor invoice may be avoidable if the campaign stops. Show those distinctions in the register. The attributed comparison can include allocated resources, while a decision about the next advertising dollar needs the costs that would actually change.

This example excludes unallocated existing corporate overhead, financing costs and income tax. Any new cost required by the campaign belongs in the relevant decision scope. A real business should agree the boundary with its finance lead or accountant. “Total” in this guide always means total within the stated scope; it never means every cost in the company.

Work through the fictional ROAS-to-ROI reconciliation

The following numbers describe an invented retailer only. The hypothetical order population, cut-offs and allocations remain the same in every row. This is how the measures can be placed side by side fairly: retain one underlying dataset, then show exactly which costs each denominator uses. ROAS and ROI do not acquire identical denominators merely because they appear in the same table.

In this fictional case, gross-revenue ROAS is 30,000 ÷ 6,000 = 5.00×, equivalent to 500% when expressed as revenue divided by media cost. Net-revenue ROAS is 27,000 ÷ 6,000 = 4.50×, or 450%. Neither percentage is a profit percentage. Each includes the revenue needed to cover costs as well as any remaining contribution.

The fictional total in-scope cost ROI is 1,800 ÷ 25,200 = 7.14%. The separately labelled marketing-cost contribution return is 1,800 ÷ 9,000 = 20.00%. Both use the same CAD 1,800 numerator. Their difference is entirely the denominator: one includes fulfilment costs and the other does not.

The fictional contribution before marketing is 40% of net revenue: 10,800 ÷ 27,000. After marketing, CAD 1,800 remains to help cover excluded costs and any residual business profit. Calling that amount “net profit” would overstate what the model establishes. The business has not yet accounted for all company expenses, and attribution has not demonstrated that advertising caused the whole amount.

Write the formulas so another person can reproduce them

Use G for gross attributed revenue, F for refunds, R for net revenue, V for net variable fulfilment costs, A for media spend and O for other included marketing costs. Then M is total marketing cost, T is total in-scope cost and P is contribution after marketing. These definitions should travel with the spreadsheet and any exported report.

• R = G − F.
• M = A + O.
• T = V + M.
• Contribution before marketing = R − V.
• P = R − T.
• Gross-revenue ROAS = G ÷ A, when A is greater than zero.
• Net-revenue ROAS = R ÷ A, when A is greater than zero.
• Total in-scope cost ROI = P ÷ T, when T is greater than zero.
• Marketing-cost contribution return = P ÷ M, when M is greater than zero.
• Contribution margin before marketing = (R − V) ÷ R, when R is greater than zero.

Keep percentage outputs as decimal ratios inside the workbook and apply percentage formatting. A hypothetical ROI of 0.07142857 displays as 7.14%. Multiplying the stored value by 100 and also applying percentage formatting would produce an incorrect display. ROAS should use a multiple format, with a separate percentage only if the report clearly labels it.

Include two arithmetic controls. Net revenue should equal variable costs plus marketing costs plus contribution after marketing. The headline ROI should also equal net revenue divided by total in-scope costs, minus one. Those identities should reconcile before rounding. A difference indicates a formula or reference problem, not a commercial insight.

These controls cannot detect an omitted invoice or prove that the revenue belongs to the campaign. They verify internal arithmetic only. A second reviewer still needs to trace the values to supporting records and examine the population, allocation choices and exclusions.

Combine amounts before combining ratios

When several campaigns belong in one report, add their compatible revenue and cost amounts first, then calculate the combined ratios. A simple average gives a small campaign the same influence as a large one. That can distort the portfolio result even if each individual calculation is correct.

Consider an entirely separate hypothetical illustration. Campaign A has CAD 1,000 media spend and CAD 6,000 net revenue, giving 6.00× ROAS. Campaign B has CAD 9,000 media spend and CAD 18,000 net revenue, giving 2.00×. The simple average is 4.00×, but the combined revenue is CAD 24,000 against CAD 10,000 media spend: the portfolio ROAS is 2.40×. The larger campaign determines more of the actual result.

The same rule applies to total in-scope cost ROI: sum the contributions after marketing and divide by the summed in-scope costs. Before combining anything, verify that campaigns share compatible currencies, outcome definitions and cut-offs. Reconcile shared expenses across the group. If each campaign includes the full creative invoice, aggregating the reports compounds the error.

Overlapping attribution also prevents a clean sum. If several reports claim the same orders, establish a mutually exclusive allocation or keep the reports separate. A portfolio reconciliation should explain how its revenue reaches the business total; arithmetic averaging cannot resolve incompatible credit rules.

Treat refunds as revenue adjustments with separate cost consequences

A returned order changes revenue, but its cost consequences depend on what actually happened. Inventory might be recovered, shipping might be unrecoverable, a processing fee might remain and a return-handling charge might be added. Reducing revenue and automatically reducing every variable cost by the same percentage can hide those differences.

In the fictional base case, CAD 16,200 is the net variable-cost amount after recognized recoveries and return-related costs. The CAD 3,000 refund is deducted from revenue once. It is not added again as an expense. A return-handling fee is a separate cost, while the reversal of an inventory cost is a separate recovery; neither should be disguised as the revenue refund itself.

Check whether the starting revenue export already includes refunds. If it does, adding the refund schedule and subtracting it again understates revenue. A reconciliation must declare whether it starts from gross revenue with a separate refund line or from already-net revenue. This model accepts only the first structure. Reconcile an already-net export back to its supported gross revenue and refund amounts before entry; do not simply relabel net revenue as gross.

Google’s documentation on the impact of conversion adjustments describes comparing original conversion values with later adjustments. That supports checking whether reported values have changed; it does not establish that a particular business’s returns have been uploaded or processed. The business ledger and the advertising report need their own reconciliation.

Show the refund risk under an explicit cost assumption

In a hypothetical stress case, increase refunds from CAD 3,000 to CAD 6,000 while holding the base case’s CAD 16,200 variable costs and CAD 9,000 marketing costs unchanged. Net revenue becomes CAD 24,000, net-revenue ROAS becomes 4.00× and contribution after marketing becomes a loss of CAD 1,200. Total in-scope cost ROI becomes −4.76%.

That scenario assumes no additional cost recoveries. It is deliberately different from a scenario where variable costs remain a fixed share of net revenue. Label the assumption beside the result. A sensitivity sheet should make disagreements about cost behaviour visible rather than produce a reassuring number from an unexplained margin setting.

Connect ROAS to contribution margin without confusing the two

Revenue buys the opportunity to cover costs; margin determines how much remains available. Two campaigns with identical net-revenue ROAS can have different returns if their product mix, fulfilment costs or marketing allocations differ. Even within one retailer, a campaign selling a different mix of items may need a different contribution assumption.

Do not take a gross-margin percentage from an accounting report and assume it equals the contribution margin required here. BDC defines gross profit margin using net sales and cost of goods sold. Our contribution calculation deducts the variable costs specified in the register, potentially including items recorded elsewhere in the accounts. Reconcile the cost categories before importing a margin.

In a hypothetical lower-margin version of the base case, retain CAD 27,000 net revenue and CAD 9,000 marketing costs, but assume a 30% contribution margin before marketing. Variable costs become CAD 18,900, leaving CAD 8,100 before marketing and a CAD 900 loss afterwards. Net-revenue ROAS remains 4.50×, while total in-scope cost ROI becomes −3.23%.

A hypothetical higher-margin version, using 50% contribution before marketing with the same revenue and marketing costs, leaves CAD 4,500 after marketing. Its total in-scope cost ROI is 20.00%. These are constructed scenarios, not a suggested margin range. Their purpose is to show why revenue efficiency alone cannot describe the economics.

Use either detailed costs or a margin assumption

The companion specification separates an actual-cost mode from a scenario mode. Actual-cost mode takes V from a reconciled cost schedule. Scenario mode derives V from net revenue and an explicitly hypothetical contribution-margin assumption. The two modes must not be added together. Otherwise the workbook subtracts variable costs once through the margin and again through the cost inputs.

Actual-cost mode can also reveal a negative contribution before marketing. Preserve that result for investigation. A scenario input of zero contribution margin has a different consequence: no amount of revenue covers positive marketing costs under that unchanged margin assumption. The workbook should say that plainly instead of showing a division error or an arbitrary very large target.

Calculate the break-even revenue ROAS for this scope

Let m be contribution margin before marketing, expressed as a decimal. If that margin remains constant, contribution before marketing is m × R. The in-scope break-even point occurs when that contribution equals total marketing cost, A + O. This derives a threshold for the specific assumptions, not a universal “good ROAS”.

Break-even net-revenue ROAS = (1 + O ÷ A) ÷ m, provided media spend and margin are positive. The hypothetical base case gives (1 + 3,000 ÷ 6,000) ÷ 0.40 = 3.75×. At that threshold, net revenue would be CAD 22,500 and contribution after marketing would be zero, under the constant-margin assumption.

The familiar shortcut of one divided by margin only applies when other marketing costs are zero within the chosen scope. For this fictional example, that shortcut would produce 2.50× and omit CAD 3,000 of included marketing costs. Both formulas are simple; only one reflects this model’s boundary.

If a report uses gross revenue and refunds are assumed to equal a constant fraction f of gross revenue, divide the net-revenue threshold by 1 − f. At the fictional 10% refund assumption, the gross-revenue break-even threshold is 3.75 ÷ 0.90 = 4.17×, rounded. It is higher because some initially attributed revenue will not be retained.

Those thresholds assume the margin, refund behaviour and marketing costs stay as specified. They do not prove that revenue will increase when spend increases. They also exclude the corporate costs already identified. Break-even in this reconciliation means zero contribution after the included costs, not complete business break-even.

A target ROI needs its own equation

For a chosen total in-scope cost ROI target t, let v be the variable-cost share of net revenue, equal to 1 − m. Required net revenue is (1 + t) × M ÷ [1 − (1 + t) × v]. The denominator must be positive. If it is zero or negative while marketing costs are positive, there is no finite positive revenue solution under the unchanged assumptions.

With the fictional base case’s CAD 9,000 marketing cost and 60% variable-cost share, a hypothetical 10% total-cost ROI target requires about CAD 29,117.65 net revenue. That corresponds to about 4.85× net-revenue ROAS. This is an algebraic scenario, not a recommended bid target. A marketing-cost contribution-return target would use a different equation because its denominator is different.

Attributed return is not incremental causal return

Attribution assigns credit to observed or modelled outcomes according to a measurement method. Incrementality asks how outcomes differ from what would have happened without the advertising. A returning customer may buy after an ad interaction while also having an existing relationship with the business. A report can assign credit without answering the counterfactual question.

Google describes Conversion Lift as a controlled comparison between exposed and control groups, and notes that availability is limited. That is a different evidence task from dividing attributed revenue by costs. A reconciliation workbook cannot turn attribution into a controlled experiment by adding a percentage called “incrementality”.

For planning, an explicitly hypothetical sensitivity can still help expose dependence on an assumption. Suppose only 75% of the fictional base case’s net attributed revenue were incremental, variable costs moved at 60% of that revenue, and all CAD 9,000 marketing costs remained incremental to the decision. The constructed incremental revenue would be CAD 20,250, variable costs CAD 12,150 and contribution after marketing a CAD 900 loss. That is a thought experiment, not a measured lift result.

The assumptions about costs are as important as the assumed revenue fraction. Applying a revenue haircut while leaving every attributed fulfilment cost unchanged answers a different question. Reclassifying an allocated salary as an avoidable expense also changes the decision. A causal evaluation needs an appropriate counterfactual for both revenue and costs, with uncertainty stated.

Keep measured incremental results, attributed reporting and hypothetical sensitivities in separate fields. Do not sum different platforms’ attributed revenue as if their claims were mutually exclusive. First reconcile them against the business’s order population and document any overlaps. Where transaction-level matching is unavailable, report the unresolved gap rather than asserting that every difference is duplicated revenue.

Use the reconciliation worksheet as an evidence record

Download the ROAS vs ROI reconciliation workbook (XLSX, fictional worked example).

A useful worksheet does more than display formulas. It keeps the reporting policy, input evidence and exceptions close enough to the result that another person can review them. The companion specification uses separate areas for scope, revenue, costs, the reconciliation, sensitivities and checks. Its sample entries are fictional and should remain labelled when copied into presentations.

Begin with the scope fields. Enter the business unit or campaign group, cohort definition, currency, time zone, interaction dates, outcome cut-off and refund cut-off. Record the source’s reporting basis and whether values are revenue, profit estimates or lead estimates. If these fields are incomplete, the result is incomplete even when the arithmetic can technically run.

Next, reconcile the inputs. Revenue needs a known population and refund treatment. Media cost needs a matching campaign scope. Variable costs need the same order population. Management, creative and measurement need documented allocations. Use one source identifier per underlying cost record and a separate allocation identifier for each assigned share. Make the allocation totals visible across all destinations. A source record split between categories still has one source identifier.

Then read the bridge in dollars before reading the ratios. Does net revenue less variable costs leave a plausible contribution? Do the included marketing amounts reconcile to their source schedule? Does contribution after marketing explain the ROI numerator? A percentage should summarize that bridge, not replace it.

Handle zero, negative and missing inputs deliberately

Zero media spend means ROAS is undefined, even when revenue exists. Show “not applicable: zero media spend”, not infinity. Zero total in-scope cost makes this ROI undefined. Zero revenue with positive costs is valid and produces a total-cost ROI of −100%. A missing cost is unknown, not zero; the workbook should stop the headline calculation until its treatment is confirmed.

Negative results also need classification. A loss is a valid output. Refunds exceeding gross revenue may be valid in a calendar-period ledger that contains older returns, but they break this simple matched-cohort input policy unless the underlying population is reconciled. Negative expense entries may represent credits. This design records a positive allocated charge and its documented credit separately, then calculates the net cost. If the source amount already reflects that credit, do not deduct it again. A credit exceeding the allocated charge requires manual reconciliation rather than an unexplained negative cost.

A time-window mismatch should stop comparison, not merely turn a cell amber while a headline ROI remains visible. Display the reason and the source fields that disagree. The same applies to duplicate cost IDs, an already-net revenue input combined with a second refund deduction, or two competing methods for calculating variable costs.

Turn the reconciliation into a decision with stated limits

Review the fictional base case alongside the refund and margin stresses. Its 5.00× gross-revenue ROAS does not prevent a loss when retained revenue falls or fulfilment costs rise. The reconciliation locates the dependence: revenue retention, contribution margin and included marketing costs. It does not identify which campaign setting should change.

Assign the next action to the uncertainty that matters. If refunds are unreconciled, the owner is the person who can verify order adjustments. If creative costs are duplicated, fix the register. If the decision depends on future repeat purchases, create a separate customer-cohort model with its own horizon and costs. Extending the revenue horizon without extending the cost horizon makes the original comparison less useful.

Also distinguish average from marginal performance. This worksheet describes the selected cohort at its recorded spend. It does not predict the return on an additional budget increment. Extra spend might reach different customers, require more creative or exceed delivery capacity. Any expansion case needs its own incremental assumptions and a review point.

Cash capacity deserves a separate check. Advertising and inventory may be paid before customer cash arrives, and refunds may arrive after the reporting period. A positive contribution figure cannot demonstrate that the business can finance the timing gap. Keep a cash schedule beside the economic reconciliation when timing affects the decision; do not change the ROI definition to make it stand in for cash flow.

For broader campaign-improvement topics, Canada Create’s PPC ROI improvement guide covers that separate task. For the factors behind advertising budgets and cost discussions, the Google Ads cost planning guide provides related context. This reconciliation remains focused on explaining the numbers under a consistent scope.

Explain differences without forcing agreement

Build a short difference schedule when an advertising export and an order ledger disagree. Start with the source total, then identify documented adjustments for date basis, refunds, excluded conversion actions, currency and attribution scope. Keep any unexplained balance on its own line. A “miscellaneous adjustment” that exists only to force a match defeats the purpose of reconciliation.

For currency, retain the original amount, original currency, conversion rate, rate date and translated amount. Select a consistent rate policy with the finance owner. Do not mix Canadian-dollar costs with US-dollar revenue merely because both exports use a dollar symbol. The fictional example avoids this complication by using CAD throughout.

Apply the same discipline to sales tax. This model excludes collected sales tax from revenue and uses costs on a documented tax basis. It does not determine whether a business can recover tax paid on purchases. The finance owner must establish that treatment; an unexplained mix of tax-inclusive costs and tax-exclusive revenue changes the result without changing campaign performance.

Finally, distinguish an estimate from an unresolved difference. An agreed accrual for an invoice not yet received can be labelled and revised. An unknown management fee should remain missing. Record the estimate’s owner, basis and expected replacement date so a provisional assumption cannot quietly become an apparent fact in the next report.

What to confirm before relying on the result

A reviewer should be able to recreate the fictional base case from the inputs without guessing any definition. For a real business, the same standard applies, with supporting records replacing the fictional values. Review the cost classification and period policy as carefully as the cell arithmetic. A perfectly functioning formula can still measure the wrong population.

• Revenue has one clear meaning, with discounts, refunds, delivery charges and tax treatment recorded.
• Media, outcomes and fulfilment costs refer to the same cohort and observation policy.
• Management, creative and measurement allocations are documented and counted once.
• The numerator and denominator appear beside every ROI result.
• Missing information, zero denominators and mismatched windows produce visible exceptions.
• Sensitivity assumptions are labelled hypothetical, and attributed results are not described as proven causal profit.
• Excluded costs and the limits of the decision remain attached to any exported chart or summary.

The formulas and fictional examples here are an educational reconciliation model. A finance lead or accountant should review the scope and cost treatment before the model supports a business commitment. Public platform documentation was checked on 8 October 2026; individual reporting configurations and data quality still require separate verification.

For help connecting your advertising reports to a documented revenue-and-cost reconciliation, discuss your measurement requirements with Canada Create.