Performance Marketing in E-Commerce — ROAS, MER and GA4 Attribution

Performance marketing is an advertising model where you pay for a specific, measurable outcome — a click, a lead, a sale — rather than for exposure. The model doesn't guarantee the outcome; it only guarantees that you pay for a defined event, not for an impression. Whether that event pays off is settled by measurement: how you calculate ROAS, how it differs from MER, which attribution model you use to split credit in GA4, and how much of it you can see at all when some users don't give consent.

Every channel covered separately in this section — Google Shopping, Meta, TikTok, price comparison sites, affiliate marketing — has its own ad platform and its own conversion-attribution logic. You still judge each one by the same four concepts: ROAS, MER, attribution and measurement under consent.

How performance marketing differs from reach advertising

The difference lies in what you pay for. Reach advertising is usually billed on exposure — a CPM (cost per thousand impressions) rate charges you whether or not anyone reacts. Performance marketing moves the billing point closer to the user's action, and it does so in several ways, not one:

• CPC (cost per click) — you pay per click, whether or not it converts afterwards. This is still billing on a proxy for the outcome (traffic), not the outcome itself (a sale).
• CPA/CPL (cost per acquisition/lead) — you are billed for, or optimise towards, the cost of a specific event: a purchase or a captured contact. Closer to the business outcome than CPC, but with no control over its value — a low-value lead and a high-value lead count the same.
• Bidding on conversion value — the ad platform (Google Ads, for instance) doesn't just get a "converted/didn't convert" signal; it gets the value of that conversion too, and sets bids to maximise that value against spend. You are still billed per interaction — under Target ROAS, Google Ads sets maximum CPC bids — but the bids are chosen for conversion value. This is the mechanism Target ROAS runs on.

The choice depends on what you want to optimise: traffic, the number of events or their value. The mechanisms don't exclude each other, and they often follow one another. Target ROAS in Google Ads requires a minimum number of recent conversions for most campaign types, so an account without that history usually starts on a simpler strategy and moves to value-based bidding only once it has collected enough events. The specific thresholds are in the section on Target ROAS.

Without measurement, each of these mechanisms bills you for an event whose profitability you can't judge. The rest of this article is about calculating that value and comparing it across channels.

ROAS — what it is and how to calculate it

ROAS (Return on Ad Spend) measures the efficiency of a single ad, campaign or channel. Google Ads Help gives the formula in the context of Target ROAS bidding: ROAS = conversion value ÷ ad spend, expressed as a percentage. Google's own worked example: "$5 USD in sales ÷ $1 USD in ad spend x 100% = 500% target ROAS" — so 500% means €5 (or any other currency) of conversion value for every €1 spent on ads.

To calculate your own campaign's ROAS, you need two numbers from the same period and the same channel: the total conversion value attributed to that campaign, and the total ad spend in that channel. Divide the first by the second and multiply by 100%. The formula is the same in Google Ads, Meta and every other paid channel.

Two things are easy to get wrong when reading ROAS off an ad platform's dashboard rather than calculating it yourself:

• The number format. Google Ads expresses target ROAS as a percentage ("500%"), but the same ratio is often given as a multiple ("5x") — the way Shopify gives MER ("4") in the section below. 500% and 5x mean exactly the same thing.
• Period and channel have to match. ROAS calculated from one period's spend and a different period's conversion value (for instance, conversions that "mature" later than the spend-reporting window) produces a number that doesn't reflect that period's real efficiency — the same risk applies to comparing ROAS between channels that use different attribution windows; we come back to this in the section on common measurement mistakes.

ROAS is not ROI. ROAS compares conversion value (revenue) against ad spend — it doesn't account for margin, product cost, logistics, or any other cost of running the store. ROI (Return on Investment) subtracts the full cost from revenue, not just the ad cost, so a campaign with a high ROAS can have a low or negative ROI if the margin on the product sold is thin. The mechanism is simple: if conversion value is rising mainly because you're selling more low-margin products, that campaign's ROAS can look good while the actual profit from those same transactions stays flat or falls. How to calculate margin per order, along with the definitions of AOV, GMV, CAC and LTV, is in our article on e-commerce KPIs.

MER — a whole-business metric

MER (Marketing Efficiency Ratio) takes the same logic from the campaign to the whole business. Per Shopify's blog, "marketing efficiency ratio (MER) is a blended marketing efficiency metric that compares total revenue with total marketing spend," and the formula is: MER = total revenue ÷ total marketing spend. Shopify's own example: $200,000 in revenue divided by $50,000 in marketing spend gives a MER of 4 — the business earned $4 for every $1 spent on marketing. Shopify draws the scope distinction directly: "ROAS helps evaluate specific ads or campaigns. MER shows whether total marketing spend is producing enough revenue across the business."

MER is an industry convention, not a metric you'll find in Google Ads or Meta Ads Manager — you calculate it yourself, from your own revenue and spend data. Shopify, as an e-commerce platform vendor, describes it on its own blog — also under the name "blended ROAS" — not as a standard set by any measurement authority.

You define the MER denominator yourself — a practical difference from ROAS, which per channel covers only the ad spend in that channel. Shopify recommends including paid ads, influencer fees, creative production, marketing tools, and agency/contractor costs in MER spend — and applying the same categories every time, because MER changes depending on whether it covers media alone or also tools, creative, agencies, contractors and salaries.

ROAS and MER can diverge, and that's a mechanism, not an anomaly: a campaign can have excellent ROAS while the whole business's MER stays flat or falls, if other channels (organic, direct, other paid) lose share in the same period that the paid channel only partly offsets. The reverse happens too — a single campaign can have weak ROAS while the whole company's MER rises, if that campaign brings in new customers whose value only shows up in later purchases, not the first transaction. ROAS tells you about one channel; MER tells you whether the sum of all channels together makes financial sense — neither one replaces the other.

Attribution and attribution models in GA4

Before you calculate ROAS or MER, you have to decide which channel gets credit for a conversion — that's what an attribution model is for. Google Analytics Help states that GA4's attribution reports today offer exactly 3 models:

• Data-driven attribution — "distributes credit for the key event based on data for each key event. It's different from the other models because it uses your account's data to calculate the actual contribution of each click interaction." Every such model is specific to each advertiser and each key event.
• Paid and organic last click — "ignores direct traffic and attributes 100% of the key event value to the last channel that the customer clicked through... before converting."
• Google paid channels last click — "attributes 100% of the key event value to the last Google Ads channel that the customer clicked through before converting." If there's no Google Ads click on the path, the model falls back to paid and organic last click.

The older rule-based models — first click, linear, time decay, and position-based — "are no longer available as of November 2023."

A model changes how credit is split, not the facts. The same set of clicks and conversions, run through different models, assigns different value to different channels, because each model splits credit for the same customer path differently. A channel that typically shows up early in the path (a generic product ad, say) looks weaker under last-click than under data-driven attribution, which credits its contribution even without the final click. Changing the attribution model in a GA4 report doesn't change what actually happened — it only changes how GA4 splits the credit for events that already occurred.

For the same reason, reports from different platforms don't add up. Google Ads, Meta and GA4 each count conversions using their own attribution logic, so the same purchase event can get attributed to different channels in different reports — or counted by two platforms at once. Adding together "conversions" reported separately by Google Ads and Meta, then comparing that sum against the number of transactions in GA4, won't give you a correct figure: that's not a measurement error, it's the result of each platform applying its own attribution model to its own data, not to one shared set of events.

Measurement under consent

Attribution only counts what reaches the measurement system, and that depends on user consent. Google Ads Help, in an announcement dated 18 January 2024, puts it plainly: to preserve ad-personalisation performance for consumers in the European Economic Area (EEA), advertisers needed to send verifiable consent signals to Google, and the change required action "before March 2024." Google describes this as upgrading the consent mode API with two new parameters — the industry calls this set of changes "v2," though Google's own page doesn't use that name.

Consent mode runs in two variants, with a different effect on what shows up in your reports. Google Ads Help describes them this way: in the basic version, Google tags are blocked until the user interacts with the consent banner, and before consent is given, no data is sent to Google at all; conversion modelling then relies on "a general model" (less granular). In the advanced version, tags load immediately with a default "no consent" state, and on refusal they send consent state and cookieless pings to Google — which lets modelling use "an advertiser-specific model" (more granular). Reporting the uplift from modelling requires consent mode to have been running for at least 7 full days, and figures only appear once a given data slice crosses a minimum data threshold.

On Meta, the Conversions API (CAPI) plays the same role on the server side. Instead of, or alongside, events collected in the browser by the Meta Pixel, events are sent by the advertiser's server, shop platform, mobile app or CRM. Per Meta for Developers, server events "are processed like events sent using the Meta Pixel, Facebook SDK for iOS or Android, mobile measurement partner SDK, offline event set, or .csv upload" — meaning they "may be used in measurement, reporting, or optimization in a similar way as other connection channels." We cover the general consent mechanism and how a consent banner affects measurement separately in our article on consent mode — here we focus on the consequence for measurement: part of the conversions in a report are directly measured, and part are modelled from the data that did give consent, and neither category is the same thing as the "actual number of transactions" visible in your own sales system.

For ROAS and MER calculated directly from your own sales data (not from an ad platform), this problem doesn't exist — revenue in your own order system is measured regardless of cookie consent. It only diverges once you compare that figure against the conversion value reported by Google Ads or Meta, because the ad platform only sees the share of traffic that gave consent or was modelled — and that's a separate reason, on top of attribution, why the sum of conversions reported by ad platforms won't necessarily match the transaction count in your store.

Target ROAS in practice

Target ROAS (tROAS) is a Smart Bidding strategy in Google Ads: you set the target, and the system sets bids itself by predicting the value of a potential conversion. The mechanism, per Google Ads Help: "Google Ads predicts future conversions and associated values using your reported conversion values... Then, Google Ads will set maximum cost-per-interaction (max. CPC) bids to maximize your conversion value, while trying to achieve an average return on ad spend (ROAS) equal to your target." In other words: the system doesn't hold to your target ROAS on every single auction — it works toward it as a campaign-level average.

From June 2026, only the strategy's label changes, not its behaviour: "Maximize conversion value with a Target ROAS" is being renamed to "Target ROAS," and Google states plainly that the bidding behaviour itself "remains exactly the same."

The strategy has eligibility thresholds that differ by campaign type. Google Ads Help lists, among others:

• Search and Shopping campaigns — at least 15 conversions in the past 30 days at the conversion-tracking level;
• Display campaigns — at least 15 conversions with valid values in the past 30 days across all campaigns combined (new Display campaigns no longer need conversion history);
• App campaigns — at least 10 conversions a day (or 300 in 30 days);
• Demand Gen campaigns — at least 50 conversions in the past 35 days within the campaign (10 of them in the last 7 days), or 100 in 35 days across all Demand Gen campaigns on the account.

Only conversions with a value above 0 count toward the threshold. How Target ROAS works in Google Shopping and Performance Max, where product data from Merchant Center determines what the system can bid on at all, is covered in our Google Shopping article.

Common measurement mistakes

Each of the following follows directly from the mechanisms above:

• Double-counting conversions across platforms. Google Ads, Meta and GA4 each count their own way, so a sum of "conversions" reported separately by each one can inflate the result — the same transaction is sometimes credited in two dashboards at once.
• Confusing revenue with margin. ROAS and MER are both calculated on conversion value, i.e. revenue — not margin. A campaign with high ROAS can run at a loss on a thin margin; the full definition of margin per order is in our e-commerce KPIs article.
• Inconsistent conversion value between channels. If one pixel/tag calculates conversion value including VAT, and another excludes VAT or shipping cost, ROAS and MER calculated from those two sources aren't comparable — even though the formula is identical.
• Comparing different attribution windows. The same set of interactions looks different depending on how far back a platform looks for clicks to credit to a conversion. Compare conversions from two platforms without checking their windows and it's easy to pick the wrong channel as the better performer.
• Judging a top-of-funnel channel by the same ROAS as a brand channel. A channel that reaches the customer early in the path will, by nature, convert less often within the same window than an ad on a branded query. A lower ROAS doesn't necessarily mean it's underperforming — only that you're measuring it by the yardstick of a channel closer to purchase, although it plays a different role.

Checklist: before you increase budget

Before you raise ad spend, check whether the measurement you're basing that decision on is reliable:

• Is conversion value calculated consistently (the same VAT and shipping-cost definition) across every channel you're comparing?
• Do you know which attribution model each platform you read a conversion count from is using?
• Do you have consent mode implemented (basic or advanced), and do you know which share of your conversions is measured versus modelled?
• Do you keep a campaign's ROAS apart from the whole business's MER, rather than treating them as one number?
• Are you calculating ROAS on revenue, and making the budget decision with margin factored in, not just revenue?
• Are you judging top-of-funnel channels by a different yardstick than channels close to purchase, instead of comparing them all against the same target ROAS?

None of these answers needs a new budget, only some order in what you already have. If you answer "I don't know" to any of them, the problem is the measurement setup, not the budget: GA4, server-side GTM and the Meta Pixel with Conversions API have to be configured correctly first for their numbers to mean anything. We describe that kind of setup, coordinated with campaigns run by a partner agency, in our online marketing and branding offer.