For ecommerce on Google, margin is the number to track
Every piece in this ecommerce series has landed somewhere unexpected. Not on media buying. On the inputs the media buyer used to control. Creative on Meta. The product feed on Google. Four conversion rates on the website.
This piece is on the fourth input, and it is the one everything else runs on. Your conversion tracking.
Because none of the previous three land if the number you send Google at the end of the funnel is the wrong one. And in most ecommerce Google accounts we audit, it is.
Every decision in your account runs on your tracking
Google Ads is a machine. It takes an input from you and outputs an auction bid. That input is a conversion signal. Every part of the account that seems to be about you making a choice, bid strategy, budget allocation, campaign structure, negative keywords, is actually being made by the algorithm on the basis of what you are telling it a conversion is. Smart Bidding is only as good as the conversion data you feed it (Google).
Which means the ceiling on your entire ecommerce Google account is the quality of your conversion tracking. Bad tracking means Smart Bidding is optimising on a fiction. Good tracking on the wrong metric means Smart Bidding is optimising on the wrong outcome. Most ecommerce accounts have one of those two problems.
Fix that, and the whole account moves. Nothing else will.
What Google actually sees when a sale happens
By default, when someone buys from your ecommerce store, your GA4 ecommerce implementation sends a purchase event to Google with a value equal to the order total (Google). Order total is revenue. It is what the customer paid.
That is the number Google Ads gets. That is the number your Target ROAS is being calibrated against. That is the number Smart Bidding is trying to maximise.
Here is the problem. Revenue is not what you actually want to maximise. You want to maximise what is left after the cost of goods, the payment processing fee, the shipping cost, the returns and the refunds. That number is the gross margin per order, and it can look nothing like the revenue number.
A 4x ROAS on a $200 order with a 15% gross margin is $30 of contribution. Google thinks it did great. Your P&L barely notices. A 2.5x ROAS on a $200 order with a 55% gross margin is $110 of contribution. Google thinks the second one is a failure. Your P&L rings the bell.
The account is scaling the first one. It should be scaling the second one.
Send margin. Not revenue.
The fix is a single change. Instead of sending the purchase event with the order total as the value, send it with the gross margin per order. Google is then bidding on the number that actually matters to your business.
You do this at the point where the purchase event fires, usually inside GA4 or the server-side tag manager. At the moment of purchase, your ecommerce platform knows the products sold, their COGS and the applicable margin. You subtract cost of goods from the order total and pass that number into the value parameter of the conversion event, instead of the revenue number.
The account will immediately look worse in the dashboard. ROAS will drop. Conversion value will drop. Anyone reviewing the account without context will panic.
Ignore them. The account is now scaling on profit, not on top-line revenue. Ninety days in, spend will drop on the campaigns and products that were burning money at high ROAS. It will climb on the ones that were flat on ROAS but printing margin. And revenue in your bank account will move without the ad platforms even understanding why.
This is the closest thing to a single-lever transformation there is in ecommerce Google Ads.
The infrastructure that lets this work
Sending margin instead of revenue is trivial technically. It only works if the tracking itself is clean. That means three specific things need to be in place.
Server-side tagging via Google Tag Manager. Client-side tags fire from the browser, which means they get blocked by ad blockers, throttled by ITP on Safari and silenced by consent refusals. Server-side tagging bypasses all of that by sending the conversion event from your server to Google's servers directly (Google). It also loads faster, protects first-party data and gives you full control over what gets sent. For any ecommerce account spending five figures a month, server-side is now table stakes.
Enhanced Conversions. Even with server-side tagging, cookies are dying, and Google will lose the ability to attribute a portion of your conversions unless it has a way to match them back. Enhanced Conversions solves this by taking hashed first-party data at the point of purchase, usually the email address, and sending it to Google alongside the conversion event (Google). Google then matches that hash against its own logged-in users on Gmail, YouTube and Chrome, and reclaims attribution for conversions that were previously invisible. Server-side implementations of Enhanced Conversions have become the standard for accurate attribution in 2026, and any account not running one is bleeding conversion data (Stape).
Consent Mode v2. If you have any EU, UK or comparable market traffic, Consent Mode is now legally required. Practically, it lets Google model conversions from users who did not consent to cookies, so the account does not go dark on privacy-conscious traffic (Shopify). Without it, the modelled conversions are lost, and the account starts to bid as if those buyers never existed.
Three layers. Each one closes a gap the layer above it cannot. Each one makes the margin-based conversion tracking downstream more accurate.
What people consistently get wrong
The same patterns show up across nearly every ecommerce Google account we audit.
They send revenue and call it done. Nine in ten accounts we audit are sending the order total as the conversion value. Smart Bidding is being asked to maximise revenue. The business is trying to maximise profit. The two only agree by accident.
They set up Enhanced Conversions once and never verify. Enhanced Conversions is not a switch. It is a data pipe that has to be tested, monitored and re-tested after every site release. A silent break can cost you 30% of your attribution for months and nobody notices until performance collapses.
They run client-side tracking on a five-figure account. If your Google Ads spend is over $10k a month and your tracking is entirely browser-side, you are paying for data that never arrives. Server-side is the cheapest single measurement upgrade in the entire ecommerce stack.
They forget Consent Mode on international traffic. UK and EU sales quietly disappear from the account, ROAS looks fine for domestic traffic, and international campaigns get paused because Google thinks nothing worked outside Australia. It did. Consent Mode was just not sending the signal home.
They import GA4 conversions instead of using Google Ads native conversion actions. GA4 uses last non-direct attribution when imported into Google Ads, which under-attributes Google's role in the buying journey. Campaigns get paused because they look unprofitable when they were actually profitable. Import margin-weighted purchase events directly through the Google Ads conversion action instead (Google).
Send Google the number that hits the bank
Every Google Ads account is having a conversation with the algorithm. The account structure is small talk. The bid strategy is a question. The conversion tracking is the answer. If the answer is the wrong number, or an incomplete number, or a number contaminated by a browser that stripped the cookie, the algorithm will bid on that number anyway. It has no other information. The brands that scale profitably on Google in 2026 are the ones who send Google the number that hits the bank, through a pipe that survives the browser, and who never assume the tracking is fine because it worked last quarter.
This is the gap we work in. At Dadek Digital we rebuild the conversion tracking layer of every ecommerce Google account we run so Smart Bidding is optimising on profit, not on revenue, and the account starts scaling toward the numbers that show up in the bank. If your Google Ads dashboard says you are winning and your P&L says you are not, that is almost always where the fix sits, and it is fixable.