Blog · Finance metrics and formulas
Incremental return on ad spend compares additional revenue with additional advertising cost. Work through a treatment and control example, normalize unequal groups, and separate the arithmetic from the evidence needed to claim causal lift.
Incremental return on ad spend measures additional revenue per additional advertising dollar caused by a defined advertising change. The calculation needs an estimate of what would have happened without that change, usually supported by a suitable controlled comparison. Ordinary attributed revenue is an allocation of observed sales and cannot supply that missing counterfactual by itself.
The intervention might be adding a campaign, increasing an agreed budget or exposing an eligible audience to an advertising treatment. The control represents the alternative being tested. Specify that alternative before interpreting the result: business as usual and no advertising answer different questions.
Google's Conversion Lift explanation describes controlled treatment and control groups as a way to measure incremental conversions. The principle matters here; a conversion count still needs a defensible revenue measure before it becomes incremental revenue. This article works from supplied study results and does not assume access to any particular experiment product.
The NIST explanation of completely randomized designs describes randomly assigning treatment levels to experimental units. Random assignment supports comparability in expectation. It does not guarantee perfect balance in a realized study, and the analyst still needs to review the assignment unit, exclusions, missing outcomes and any spillover between groups.
For a US company reviewing account-level advertising, the unit might be an eligible business account. Keep the complete assigned population in the comparison. Dividing only by accounts that clicked or bought selects on behavior after assignment and can destroy the original comparison.
Incremental ROAS = estimated incremental net revenue / incremental advertising spend
Incremental net revenue = observed treatment revenue - estimated treatment revenue without the tested change
The numerator is revenue, not profit. Revenue should use the same definition across groups, including the treatment of returns, discounts and the observation cutoff. Incremental spend measures the cost difference for the same intervention and population.
When the control has no relevant advertising cost, the denominator may equal the treatment advertising cost. When both groups have business-as-usual advertising, compare treatment spend with a suitable control-based spending baseline. Do not subtract unequal groups' total costs without adjusting the comparison.
The marketing ROI versus ROAS guide covers attributed return conventions. Marketing spend analytics covers the cost and outcome reconciliation that should precede this narrower causal comparison.
An illustrative US industrial distributor evaluates an added campaign above business-as-usual advertising. Its measurement owner supplies results from randomly assigned eligible accounts, observed over the same window. Assignment sizes differ, and the approved comparison uses revenue and advertising cost per assigned account. Assume comparable measurement and baseline advertising per account.
These are the supplied inputs. Revenue is net of the agreed credits and returns. The last row is a finance assumption used later to translate revenue lift into contribution; it is not estimated from the experiment totals.
| Input | Treatment | Control |
|---|---|---|
| Assigned eligible accounts | 12,000 | 10,000 |
| Observed net revenue (USD) | 132,000 | 100,000 |
| Relevant advertising spend (USD) | 24,000 | 10,000 |
| Assumed contribution margin before advertising | 40% | 40% |
Normalize the control to the treatment population. The scaling factor is 12,000 / 10,000 = 1.2. This normalization is appropriate for this account-randomized example under its stated assumptions; a geographic or clustered study may require a different approved estimator.
| Step | Calculation | Result |
|---|---|---|
| Treatment revenue per account (USD) | 132,000 / 12,000 | 11.00 |
| Control revenue per account (USD) | 100,000 / 10,000 | 10.00 |
| Treatment revenue without added campaign (USD) | 10.00 x 12,000 | 120,000 |
| Estimated incremental revenue (USD) | 132,000 - 120,000 | 12,000 |
| Treatment spending without added campaign (USD) | 10,000 x 1.2 | 12,000 |
| Incremental advertising spend (USD) | 24,000 - 12,000 | 12,000 |
| Incremental ROAS | 12,000 / 12,000 | 1.00x |
The revenue lift relative to the estimated baseline is $12,000 / $120,000 = 10.0%. That percentage and the 1.00x return answer different questions: one compares revenue with its baseline, while the other compares additional revenue with additional advertising cost. Dollar calculations are exact here; ratios display two decimals and the lift percentage one decimal.
Two identities prove the arithmetic: baseline revenue of $120,000 plus incremental revenue of $12,000 equals treatment revenue of $132,000. Baseline advertising cost of $12,000 plus incremental spend of $12,000 equals treatment spend of $24,000. These identities check the calculation, rather than proving that the study design identifies a causal effect.
The supplied contribution assumption changes the budget interpretation. Estimated incremental contribution before advertising is $12,000 x 40% = $4,800. Subtracting $12,000 of incremental advertising leaves negative $7,200 before any other incremental campaign costs. A positive revenue lift can therefore fail the economic test.
With the same margin and no additional costs, contribution break-even would require an incremental revenue multiple of 1 / 40% = 2.50x. The gross margin and contribution margin comparison explains why the selected cost boundary matters. Use the margin on the additional sales, rather than automatically substituting the company's average gross margin.
Now change only the control revenue input to $90,000, holding the other inputs fixed. Control revenue per account becomes $9.00, the treatment baseline becomes $108,000, and incremental revenue becomes $24,000. Incremental spend remains $12,000, so incremental ROAS becomes 2.00x. This is a sensitivity illustration, not a second observed study or a confidence interval.
The marketing measurement owner should approve the assignment method, estimator, observation window and treatment definition. Finance should approve net revenue, advertising cost and contribution scope. Business leaders use the resulting estimate to judge the tested intervention, alongside its uncertainty and operational constraints.
Group totals alone do not provide the variability needed for a defensible confidence interval. Request the study's uncertainty estimate and method from its measurement owner. If the study randomized regions, uncertainty must respect that assignment structure; treating every account as an independently randomized observation would overstate the evidence.
The association and cause guide covers the wider distinction between a pattern and a causal conclusion. Covirage is not affiliated with Google.
Bring the approved treatment and control results, study definitions and corresponding campaign cost file. Covirage maps the columns; its tools compute and check the supplied comparisons, and the external AI model explains the result with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.
See revenue driver analysis for the supporting approach to examining changes in business revenue.
Attributed ROAS uses revenue credited to advertising under an allocation rule. Incremental ROAS uses an estimate of revenue caused by the tested advertising change. The two can differ because credited customers may have purchased even without that advertising.
Compare it with the contribution margin on the additional revenue, the included campaign costs and the company's required return. A revenue multiple alone is insufficient. Also consider uncertainty and whether the measured intervention represents the proposed next use of the budget.
It can supply observed revenue, but it cannot reveal what those same customers would have done without the advertising change. Bring an approved estimate of that counterfactual, such as a suitable controlled study, together with the corresponding cost and scope definitions.