ROAS, MER or contribution margin: which to trust

Trust each one for one question only. Contribution margin sets the line a euro of ad spend has to clear, blended MER (total revenue divided by total ad spend) shows whether the whole business clears it, and platform ROAS ranks ads inside one platform. None of the three measures what the ads caused: in 15 Facebook advertising experiments, observational methods "often fail to produce the same effects as the randomized experiments" (Gordon et al., Marketing Science, 2019).

Which question does each metric answer?

Platform ROAS is the right tool for that ranking: Google's conversion-value page describes using it to "identify keywords, ad groups, and campaigns that show a high or low return on investment".

Why does platform ROAS mislead, and what did experiments find?

Three reasons, the first two from Google's own documentation:

1. The windows differ. Google Ads counts click-through conversions for 30 days by default and view-through conversions for 1 day (Google Ads Help, 30 September 2026), and other platforms choose other windows.
2. Part of the number is modelled. Google says its Conversions column "reports both modeled and observed conversions", and that modelled conversions can take up to 5 days to stabilise (Google Ads Help, 30 September 2026).
3. There is no control group. Attribution credits the orders that followed an ad. It doesn't compare them with people who never saw it.

Two peer-reviewed papers looked at how hard advertising effects are to measure. Neither audited Ads Manager's own ROAS, so read them as evidence about methods, not about one product.

Gordon and colleagues (Marketing Science, 2019) used 15 US advertising experiments at Facebook, with 500 million user-experiment observations and 1.6 billion ad impressions. Their abstract says: "The observational methods often fail to produce the same effects as the randomized experiments, even after conditioning on extensive demographic and behavioral variables."

Lewis and Rao (Quarterly Journal of Economics, 2015) analysed 25 large field experiments with major US retailers and brokerages. Their abstract says that "the median confidence interval on return on investment is over 100 percentage points wide", because individual-level sales are very volatile: "a coefficient of variation of 10 is common".

Two readings follow. A platform's ROAS is a ranking, so don't quote its decimals as returns. And even a randomised test only narrows the answer.

When does MER help, and when does it mislead?

MER counts every order once, because it divides your store's total revenue by your total ad spend, so no platform can credit itself twice inside it. It is also blunt: it moves with promotions, season, email and repeat buyers, so a better MER doesn't show the ads worked, and it can't say which channel to move. That is reasoning from the definition, not a measured finding. Blended MER vs platform ROAS has the sum.

Use MER as a cross-check on the platforms. Divide the revenue they credit by your store's revenue for the same dates. For illustration: three platforms that each credit €40,000 against €100,000 of store revenue claim 120,000 / 100,000 = 1.2 times the store's revenue, so some orders are credited twice.

Where does contribution margin fit?

It sets the floor. OpenStax, an open accounting textbook, defines total contribution margin as "the total amount by which total sales exceed total variable costs". Divide one by your contribution margin, as a share of revenue, and you have your break-even ROAS. When revenue and spend are totals, the same number is your break-even MER. For illustration: a 40% margin gives 1 / 0.40 = 2.5. ROAS vs ROI: same inputs, different answers works through a good ROAS that loses money.

How do you check all three each week?

Three lines from your own exports, for one full week, in one currency and one time zone:

1. Break-even: contribution margin per order after product cost, shipping, payment fees and refunds. Break-even is one divided by that margin.
2. MER: Shopify net sales, divided by ad spend across every platform. The order to do things in says why net sales.
3. Claim ratio: the revenue the platforms credit, divided by the Shopify revenue for the same dates.

Pass and fail:

• MER above break-even: the orders cover the ads at today's margin. Below it: they don't.
• Claim ratio of one or less: the platforms together claim no more than the store took. Above one: some orders are credited more than once, and the excess is the size of the overlap.
• After a budget change, compare direction. If a platform's ROAS holds while MER falls, that platform's number isn't tracking your revenue. Season can do this too, so repeat the check the next week before you act.

The ad platform over-reporting checker does the claim ratio. When the lines disagree, or before you cut or scale a channel, settle it with a holdout, not an argument: A 30-day ecommerce attribution playbook sets one up, and Cut a channel the right way says why.

Sources, 30 September 2026: A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook (Gordon et al., Marketing Science, 2019); The Unfavorable Economics of Measuring the Returns to Advertising (Lewis and Rao, Quarterly Journal of Economics, 2015); About conversion values (Google Ads Help, Google, 2026); About conversion windows (Google Ads Help, Google, 2026); About modeled online conversions (Google Ads Help, Google, 2026); Contribution margin (OpenStax, 2019). The euro sums are arithmetic on assumed numbers, not benchmarks.

Related answers

• Blended MER vs platform ROAS: which should drive spend?
• ROAS vs ROI: same inputs, different answers
• A 30-day ecommerce attribution playbook
• Marketing ROI measurement: the order to do things in
• Incrementality testing, explained