Creative-level ROAS: compare more than a ranking

Inspect what each creative actually received

Compare spend, impressions, clicks, completed orders, order value, and time since delivery. Check whether one creative reached returning customers while another reached new prospects. Note landing-page and offer differences as well. A visual change combined with a discount is more than a creative-only comparison.

A small winner may have fragile evidence

Illustrative example: creative A spends $100 and receives two $200 orders, giving $400 / $100 = 4×. Creative B spends $1,000 and receives 30 orders worth $100 each, giving $3,000 / $1,000 = 3×. A ranks first by ROAS, but its result rests on only two orders and a different average order value.

If one A order is later fully refunded, its reconciled revenue falls to $200 and ROAS to 2×. That is a sensitivity example, not a prediction that a refund will happen or proof that B is superior.

Compare contribution as well as revenue

Different creatives can sell different products. Calculate revenue less relevant order costs and advertising spend on a consistent basis. A higher revenue ratio can still leave less contribution if it mainly sells low-margin items or uses a costly incentive.

Use a test to answer a causal question

If you want to know whether a specific message improves results, define the change, outcome, and comparison before running an experiment. Keep other conditions as comparable as the test design permits. Avoid stopping solely at a favorable early result, and allow relevant conversions to mature.

Use ordinary delivery reports to find hypotheses and operational issues. Use a suitably designed test when the decision depends on attributing an improvement to the creative itself.

Examples are illustrative, not client results or benchmarks. The calculations use the inputs shown; real-world decisions depend on your records and assumptions. Report a correction.