Why retargeting changes the calculator's inputs
On the default calculator page, revenue from ads means the revenue the ad platform or analytics tool attributes to the campaign. That works tolerably for campaigns that reach new people, where most buyers would not have arrived without the ad. Retargeting reverses the situation. Its audience is selected because it is about to buy: shoppers who left items in a basket, visitors who viewed a product several times, customers due a repeat order. A large share of them will purchase in the following days with or without another advertisement.
Attribution cannot tell those buyers apart from the ones the ad persuaded. It sees an impression or a click, then a purchase, and gives the campaign credit. Retargeting therefore tends to post the most attractive ROAS in the account while causing a smaller share of its revenue than any other line. The explanation of prospecting and remarketing describes how this distorts budget splits between the two jobs.
The calculator's arithmetic is not wrong; the input is. Swapping reported revenue for the revenue the ads caused turns the output into incremental ROAS, the figure a budget decision needs. The spend and margin inputs change only slightly, but the margin should reflect what retargeted visitors buy, which can differ from the catalogue average.
The inputs for retargeting campaigns
| Calculator input | On the default page | For retargeting campaigns | Where to find it |
|---|---|---|---|
| Revenue from ads | Attributed revenue for the period | Incremental revenue: the exposed group's revenue minus the revenue the holdout group's rate implies for an audience of the same size | Holdout or conversion-lift test results, not the platform's attributed column |
| Ad spend | Media cost for the period | Retargeting spend only, for the test window | Campaign-level spend for the retargeting campaigns |
| Gross margin | Everyday gross margin | Margin of the products retargeted visitors actually buy | Finance margin by product line, weighted by what the audience purchased |
The revenue row carries the derivation. Split the retargeting audience at random, show ads to one group and withhold them from the other, then measure revenue per person in each over the same window. Multiply the holdout's revenue per person by the size of the exposed group to estimate what that group would have spent anyway. Subtract this from the exposed group's actual revenue. What is left is the revenue the campaign caused, and it belongs in the revenue box. The remarketing holdout experiment sets out the design in detail, and the guide to incrementality testing covers alternatives when a clean audience split is not available, such as geographic tests.
Entering the retargeting inputs step by step
Run or retrieve a holdout for the same audience
Use a test that withheld retargeting from a random slice of the same audience definition, with the same lookback window and exclusions the live campaigns use. A test on a different audience measures a different campaign.
Calculate the revenue the ads caused
Scale the holdout's revenue per person up to the exposed group's size and subtract it from the exposed group's revenue. Keep the platform's reported figure to hand for comparison, but do not enter it.
Match spend and margin to the test window
Take the retargeting spend for exactly the dates the test ran. Weight the margin by the products the exposed group bought, since retargeted baskets often lean towards items people browsed repeatedly.
Enter the inputs and record both returns
Type the incremental revenue, the spend and the margin into the calculator. Note the ROAS it returns as incremental ROAS, and write the reported ROAS beside it so the gap is visible to anyone reviewing the budget.
A worked retargeting programme with illustrative inputs
An invented hiking-gear shop tests its retargeting
Fellrunner Outdoor, an invented online shop selling hiking gear, spends $6,000 in a month on retargeting site visitors. The ad platform reports $60,000 of revenue from those campaigns, a reported ROAS of 10. The shop holds back a random slice of the audience for the same month. Scaled to the exposed group's size, the holdout's purchase rate implies that $36,000 of the reported revenue would have arrived anyway, so incremental revenue is $60,000 − $36,000 = $24,000. At a 35% margin on the products those visitors buy, the calculator returns an incremental ROAS of 4.0 against a break-even of about 2.9. Gross profit after ad spend is $24,000 × 35% − $6,000 = $2,400, against the $15,000 the reported figure would have suggested. Only 40% of the credited revenue was caused by the ads.
ROAS Calculator
Open the ROAS Calculator →Prefilled with the example's inputs: Revenue from ads $24,000 · Ad spend $6,000 · Gross margin 35% → ROAS 4× · Break-even ROAS 2.86× · Gross profit after ad spend $2,400.
The shop and all its figures are invented. Replace the revenue in the embedded calculator with the reported figure to see how far the profit line rises when credit is mistaken for cause.
Interpreting incremental ROAS for retargeting
Read the ROAS output first, and read it as incremental ROAS. It is this number, not the reported one, that belongs next to the break-even ROAS on the same screen. Incremental ROAS above break-even means the campaigns caused enough sales to pay for themselves at the stated margin; below it, the shop would be better off spending less on retargeting, whatever the platform shows.
Then measure the gap. Dividing incremental revenue by reported revenue gives the share of credited sales the ads truly caused; the remainder is the share that would have happened anyway. A small caused share is normal for audiences of basket abandoners and recent customers, and it is the clearest argument for tightening the audience rather than enlarging it.
Next, treat the profit line as conditional on scale. A retargeting audience is finite. A healthy incremental return at the current budget may not survive more spend, because extra budget mostly buys extra frequency against the same people, and each additional exposure persuades fewer of them. Before raising the budget, consider testing the higher level rather than projecting the current IncrementalityConversions caused by marketing. forward.
Finally, decide when the figure expires. The incremental return describes one audience definition, one set of exclusions and one creative at one time of year. Change any of them and the share of buyers who would have purchased anyway changes too. Re-test after such changes rather than reusing an old result, and keep the Break-even ROASMinimum ROAS to avoid a loss. comparison anchored on caused sales, not credited ones.
Pitfalls when retargeting numbers go into the calculator
- Entering platform-reported revenue. This is the most common mistake and it always flatters retargeting, because the audience was chosen for its likelihood to buy.
- Running the holdout too short for the purchase cycle. If customers usually take a couple of weeks to decide, a test that ends sooner catches the ads' quick wins and misses the holdout's later purchases, or the other way round.
- Leaving recent buyers in the audience. People who have just purchased rarely buy again at once. Their presence inflates reported revenue on repeat orders they would have placed regardless and wastes impressions.
- Comparing incremental ROAS with prospecting's reported ROAS. The two figures measure different things. Put each channel on the same basis, incremental against incremental, before moving budget between them.
- Treating one test as permanent. A holdout from a quiet month can understate or overstate the caused share during peak trading, when many more visitors arrive ready to buy.
