A ROAS of three times can cover one offer’s advertising cost and fail to cover another’s. When you buy placements in ChatGPT, revenue per advertising unit is only half the economic assessment. The other half is how much of that revenue remains after supplying what the customer bought.
This guide derives a margin-based break-even threshold. It is an editorial calculation method, not a recommended industry target or a documented platform ROAS setting. Every numerical example is hypothetical. The scope is advertising shown in ChatGPT and the sales being evaluated alongside its media cost.
Choose the revenue basis before calculating
Define the numerator first. This worksheet uses net revenue after discounts and refunds, excluding VAT or other collected sales taxes. Variable costs are deducted when calculating contribution margin. Calling contribution itself revenue would create a different ratio and make the result difficult to compare with an ordinary ROAS measure.
Check what value your purchase events actually send. A checkout total may use a different basis from net revenue in the finance records. OpenAI Conversion Tracking explains how events connect with advertising. Your reconciliation must still establish which economic revenue those events represent.
Revenue and advertising cost also need one currency. If an ad account reports spend in one currency while orders arrive in several, use a consistent conversion rule and retain its date. Adding original amounts from different currencies does not create a meaningful revenue total, even when a spreadsheet allows the operation.
Derive the ratio rather than choosing a round number
Let R mean net revenue, A mean advertising cost and m mean contribution margin before advertising, expressed as a decimal. Contribution after advertising is R × m − A. Break-even within this definition occurs when R × m = A. Dividing through gives a required R / A ratio of 1 / m.
With a 40% contribution margin, the calculation is 1 / 0.40 = 2.50. Each unit spent on advertising therefore requires 2.50 units of comparable net revenue to cover that advertising cost. At the threshold, nothing remains for fixed costs or profit. A practical operating target may need to be higher.
Always include the unit: 2.50 times or 250%. A bare number such as 250 is ambiguous. If contribution margin is zero or negative, no positive finite ROAS threshold solves the problem under this model. The offer’s economics must change, or the spending decision needs another explicitly defined rationale.
Product mix changes the answer
Consider two hypothetical product groups. Group H earns a 60% contribution margin before advertising; group L earns 20%. If each contributes SEK 50,000 of net revenue, their contributions are SEK 30,000 and SEK 10,000. Total contribution is SEK 40,000 on SEK 100,000 revenue, producing a weighted margin of 40%.
Now change the mix to SEK 20,000 of revenue from H and SEK 80,000 from L. Contributions become SEK 12,000 and SEK 16,000. The total is SEK 28,000, so the weighted margin falls to 28%. Break-even ROAS becomes 1 / 0.28, approximately 3.57 times. Averaging 60% and 20% without revenue weights would still produce 40%, which is wrong for this mix.
Suppose advertising cost is SEK 30,000 in both cases. ROAS is 100,000 / 30,000, approximately 3.33 times. The first mix retains SEK 10,000 after advertising. The second produces a SEK 2,000 shortfall. An identical revenue ratio can therefore describe substantially different economics.
This is especially relevant when a ChatGPT campaign promotes several offers or a destination allows shoppers to choose beyond the advertised item. Base the margin on the purchases included in the analysis. The margin of the featured product alone may not represent the resulting basket.
Same ROAS, different contribution
- Mix A: 40% margin
SEK 40,000 contribution before ads gives a 2.50 times break-even threshold.
- Mix B: 28% margin
SEK 28,000 contribution before ads gives an approximately 3.57 times threshold.
- Observed ratio: 3.33
After SEK 30,000 of ads, A retains SEK 10,000 while B has a SEK 2,000 shortfall.
Decide what must remain after media cost
Suppose the business wants to retain 10% of net revenue for fixed costs and profit when contribution margin is 40%. That leaves 30% for advertising, giving a working target of 1 / 0.30, approximately 3.33 times. The retained percentage is an illustrative operating requirement, not a universal safety allowance.
Specify whether the requirement is a share of revenue, an amount per order or a total for the reporting period. Those choices produce different calculations. If the decision is naturally expressed in money per purchase, break-even CPA may be more useful. Choose the measure that represents the business requirement rather than the most flattering presentation.
Check the comparison before acting
Use the same campaign scope, period and conversion selection for revenue and cost. OpenAI Reporting is the reference for supported reporting measures. Clearly label any internal adjustment that combines reported values with refunds or order records to create the worksheet’s net revenue.
Attributed revenue does not establish that every included sale was caused by the advertisement. The arithmetic can be correct while describing attribution rather than incremental impact. Do not present the gap above break-even as causal proof of additional sales without separate evidence.
Record the product mix behind the margin and the changes that require a recalculation. A new offer, deeper discount or higher refund amount can move the threshold even when advertising cost is unchanged. Keep precision in the worksheet and round only the displayed result.
Cash timing remains outside this ratio. A campaign can cover media cost under the model while collecting customer payments after its bills fall due. Use an advertising test cash reserve when the next decision concerns the money that must be available during the test.
Sources and scope
Determine the revenue per advertising unit required for a defined product mix to cover its ChatGPT media cost.
Working methods and examples are editorial suggestions. Check current platform requirements and available features before implementation.