MER vs ROAS: which number do you spend against?

Ad spend is up. Sales are flat. Shopify says one number. Meta says another. GA4 says a third. The meeting needs a spend decision by Friday. Operators almost never type “why don’t Shopify and GA4 match.” The live form is MER vs ROAS. Which efficiency story do you believe? Which one do you spend against?

MER (marketing efficiency ratio) is usually blended revenue over blended marketing spend. Platform ROAS is attributed revenue over that platform’s spend, inside that platform’s model. They are allowed to disagree. The missing part sits underneath both: contribution and incrementality on your orders. The loudest tile is not the decision.

This is not a second what is a good ROAS for ecommerce page. That piece is the judgment on 2× vs 4×. This one is the measurement argument. What each count includes. Why the stack disagrees on purpose. When incrementality should overrule both.

Do not rank this as “what is a good MER.” Bare MER is a fund-fee SERP. Keep the ecommerce frame.

What MER counts

MER asks: for all the marketing we paid for, how much revenue did the business take?

Typical construction:

numerator: store revenue (sometimes net of returns, often not)
denominator: paid media plus, if you are honest, the costs you still treat as “marketing”: affiliates, retainers, tools, creative production

A clean MER moves when the business moves. It does not care which platform claimed the order. That is the point. It is a blended sanity check.

If you exclude brand search because “that would have happened anyway,” MER gets prettier and less useful. If you exclude email because Klaviyo is “owned,” you have defined away a lever. If the numerator is gross merchandise value before returns, fashion MER in a sale week is a mood.

Use MER to answer: is the whole acquisition machine getting cheaper or more expensive relative to sales? Do not use it to pick a campaign in Ads Manager. It cannot see a campaign.

Beauty and fashion both need a returns assumption in the numerator before you call MER “the real number.” A 4.0 MER on jackets that come back at 30% is not a 4.0 MER. A 2.2 MER on a replenishing moisturiser that stays sold is a different machine.

What platform ROAS counts

Platform ROAS asks: of the spend in this ads account, how much revenue will this platform claim?

It sees clicks, views and a model. It does not see product cost, shipping, payment fees, markdown or the customer who would have bought from the email they already get. Meta, Google and TikTok each keep their own window. Shopify and GA4 will disagree with all three. That disagreement is three answers to “who gets the sale.” It is not a bug in your pixel.

ROAS is useful for operating an auction. It is a poor single number for operating a P&L. The judgment page already covers why 2× vs 4× vs 5× is not a verdict, and why a 4× tile can sit on a business that is not paying back. Do not re-litigate the ranges here. Take them as read. Then ask which dashboard you are willing to spend against when the tiles fight.

A short aside, not a textbook. ROAS vs ROI is a vocabulary problem. ROI, used carefully, is closer to profit over cost. Operators still say ROAS when they mean “is paid working.” Keep the words still. If you need profit, say contribution or payback. See unit economics for a Shopify fashion or beauty brand and what is a good CAC / LTV:CAC / payback.

Why Shopify, GA4 and the ads manager disagree on purpose

Each system is counting a different job.

Shopify is close to orders and cash. It is weak at telling you which ad caused the order. It will happily show you sales that paid never touched.
GA4 is a session and event model with its own attribution settings. It is useful for path and device. It is not the finance system.
Meta / Google / TikTok are trying to prove their spend worked, inside rules they control.

When someone asks “which number do I spend against?” they are usually asking which of these to treat as true. None of them is true in the sense of contribution. One of them is the operational number for a channel. One of them is the blended number for the company. The spend decision needs a third: did this increment happen, and was it worth it after cost and returns.

A practical rule. Not a full method:

use platform ROAS to kill obvious waste inside a channel, once you know the channel’s claim is not wildly inflated
use MER to stop a company-level lie (“every channel is 5× and we are still tight”)
use contribution and payback to decide whether to add spend

If Shopify revenue is flat and every ROAS tile is healthy, believe Shopify’s sales first and treat the tiles as suspects. If Shopify is up and MER is down, you scaled spend faster than sales. The platforms may still be “working” in their own model.

Incrementality: the sales that would have happened anyway

Incrementality asks: of the orders we attributed to this campaign, how many would have happened without it?

Brand search, retargeting and a loud beauty hero SKU are classic credit thieves. The platform shows a handsome ROAS because it saw the last click. MER may look stable because you did not actually create demand. You paid to stand in front of a purchase that email, habit or the drop calendar was already going to complete.

Fashion example: a remarketing campaign against /products/linen-shirt during a 20% off week. ROAS looks excellent. Many of those buyers had the jacket in /cart from organic or email. You paid for a discount they were going to use.

Beauty example: always-on prospecting on the hero moisturiser while the replenishment cycle is already in Klaviyo. ROAS can look fine. Incrementality can be thin. The customer was going to repurchase when the bottle ran down.

You will not get incrementality from a blog formula. You get it from a holdout, a geo split, a pause, or at least a comparison that does not give the platform the last word. The full test design is the work. The literacy point is: attributed ROAS is not proof the spend created the sale.

Incrementality testing as a slug is thin. Keep it here, as the missing layer on MER vs ROAS. Not as a ninth Thinking URL.

Fashion and beauty: different clocks, same argument

Fashion paid is often a bet on a first order against inventory with a markdown clock. Platform ROAS in week one of a drop can look heroic. MER over the life of the style can look ordinary once you clear size 14 and take 30% off. Incrementality on launch week is not incrementality on week six.

Beauty paid is often a bet on a first SKU that might replenish. Platform ROAS on a kit or gift-with-purchase can buy customers who never pay full price for the hero. MER can look acceptable in the gift week and weak in the quiet month after. Incrementality matters twice. Did the first order exist without the gift? Did the second order exist without another offer?

Neither category gets a TwoKai league table of “good MER.” A blended 3× can be healthy on a replenishing routine and reckless on seasonal outerwear. Split the same way as the ROAS and CAC pieces: first-order contribution vs a repeat window you can defend. If you cannot defend the window, do not use it to excuse a weak first order or a tile that stole brand demand.

Which number do you spend against?

Spend against the number that matches the decision you are actually making.

Kill a broken ad set this afternoon. Lean on platform ROAS plus frequency and creative fatigue. Do not lean on MER. Too slow. Too blended.
Decide whether paid is eating the company. Lean on MER, then contribution. Do not lean on a single campaign ROAS.
Add 20% spend next month. Lean on incrementality and payback on the first SKU you will scale. Do not lean on the prettiest tile in the room.
Reconcile Shopify vs GA4 vs Meta. Lean on orders and cash first, then each model’s job. Do not lean on a debate about which dashboard is “wrong.”

A tripwire set, not a diagnosis:

Spend up, sales flat, platform ROAS stable → mix, incrementality, or discounting.
ROAS up, MER down → credit stealing or channel mix.
MER up, contribution down → returns, markdown, or a basket that looks bigger than it is.
Every tile fine, second order absent → you bought a discounter. That is a retention problem wearing a measurement hat.

When the tripwire fires, the next step is not a new benchmark. It is to pick the decision metric and measure it on your orders. Reconciling the stack is the work.

What's missing

This page can tell you what MER and platform ROAS each count, why they disagree, and why incrementality sits underneath both. It cannot tell you which of your channels is claiming sales it did not cause, or what contribution looks like after ads and returns on your catalogue. That is measurement on your orders. Shopify, GA4 and the ads manager included. None of them in charge. You can check one typed order at the ROAS calculator.

If you want that investigation, Explore an opportunity with TwoKai.