Why your Google Ads ROAS is lying to you — and what to do about it

The ROAS in your Google Ads dashboard measures conversions that Google’s system can observe and attribute to a click in their network. It does not measure your actual revenue return for every euro spent.

The distinction matters enormously. Google can only see conversions that are reported back through a tracking mechanism they control — a browser pixel, a GA4 import, or a server-side conversion API. When tracking is incomplete, the denominator (cost) is accurate, but the numerator (revenue) can be dramatically wrong.

But here’s the counterintuitive part: the problem isn’t that Google undercounts. It’s that Google overcounts — through a mechanism called double-counting.

Most advertisers run both browser-based conversion tracking (the gtag.js pixel) and imported goals from Google Analytics 4. Both report to the same ad account. Both attribute to the same click.

The result: a single purchase is reported as two separate conversions — doubling the revenue signal that Google’s algorithm uses to make bidding decisions.

On top of that, Google’s default attribution window is 30 days for purchase events. A customer who clicks your ad, browses, then returns two weeks later via direct or organic search — Google still claims full credit for that conversion through view-through or assisted attribution.

Understanding exactly where the overcounting happens makes it possible to fix. There are three main sources.

The formula is deceptively simple. Getting the right data to feed it is where most advertisers fall short.

Verified revenue must come from a single source of truth — your CRM, your Shopify admin, or your payment processor. Not Google Ads. Not GA4. Those are attribution tools, not revenue ledgers.

Once you have your real revenue figure for a given period, compare it to what Google Ads reported for the same window. The ratio between them is your inflation factor — and the gap is your Attribution Gap.

Fixing inflated ROAS is a three-step process. None of these steps are difficult — but skipping any one of them leaves the problem partially intact.

Step 1: Audit and remove duplicate conversion actions

Go to your Google Ads account → Tools → Conversions. For every purchase or lead action, check whether the same event is also being imported from GA4. Keep only one source per event type. If you have both a tag-based purchase event and a GA4 goal import for the same action, disable the GA4 import.

Step 2: Switch to server-side conversion tracking with deduplication

Browser-based pixels are blocked by ~35% of users — ad blockers, iOS Safari’s ITP, Firefox Enhanced Tracking Protection. Server-side tracking fires from your own infrastructure and cannot be blocked by the browser.

More importantly, server-side events let you enforce deduplication before the conversion reaches Google. The Google Ads Enhanced Conversions API supports a transaction_id field — if you include the order ID here, Google will automatically reject any duplicate event with the same ID within 24 hours.

A minimal server-side Enhanced Conversions payload looks like this:

{
  "conversion_action": "purchase",
  "order_id": "ORD-8821",
  "currency_code": "EUR",
  "value": 249.95,
  "user_data": {
    "email": "<sha256-hashed-email>",
    "address": {
      "country": "NL"
    }
  },
  "transaction_id": "ORD-8821"
}

The transaction_id is your deduplication key. If your browser pixel and your server both fire for the same purchase, Google will merge them into a single conversion event — eliminating the double-count.

Step 3: Track the gap as an ongoing metric

Even after you fix the tracking setup, measure the inflation ratio weekly. Your verified revenue versus platform-reported revenue is a leading indicator of tracking drift — and it catches issues like a broken server-side integration before they compound for an entire month.

A reported ROAS of 6.4× that’s actually 3.8× is not a vanity problem. It means your smart bidding campaigns are training on inflated signals. It means you’re allocating budget based on fictitious performance data. And it compounds: as you shift spend to campaigns that look stronger, you’re often shifting to campaigns that are better at double-counting — not better at driving actual revenue.

The fix requires no new channels, no new budget, and no agency retainer. It requires accurate measurement infrastructure — which is exactly what server-side tracking with proper deduplication provides.

Fix the measurement first. Then let the algorithm do its job on clean data.