Shopify for orders and revenue, GA4 for behavior and sources. One tool per question, and never both in the same calculation. A colleague summed it up in a meeting, talking about two tools that will never agree: "they'll never count the same way, so don't mix them in the same analysis."
The benchmark
Shopify records every order: it's your cash register. GA4 only sees visits that agreed to be measured and whose browser lets tracking through, then it fills part of the gap with modeling. So it often misses purchases, and sometimes it invents them.
At a ready-to-wear brand we audited in May, GA4 counted 1,026 purchases in one month against 568 Shopify orders, almost double: two pipelines each sent the purchase event, and neither knew the other existed. Here, nothing was missing. There were duplicates.
So once a month, calculate your coverage rate: purchases seen by GA4 divided by Shopify orders. Above 90%, there's nothing to do. Between 70 and 90%, trends remain readable.
Below 70%, fix your tracking before commenting on anything. Above 100%, a tag is sending the purchase twice.
The rule
In front of leadership, each number has a single source. Revenue comes from Shopify. The breakdown by channel comes from GA4, presented as a trend, not as a total.
The steps
- The mapping table. Write down in black and white which source answers which question, one line per decision.
- The coverage rate. Calculate it and track it every month, with the date of each banner or tag change next to it.
- The gap, shown openly. One line in the report is enough: "GA4 sees 78% of our orders this month."
- The forbidden calculation. Never recalculate a conversion rate by mixing GA4 sessions and Shopify orders.
- The line for leadership:
"Revenue comes from the register. The breakdown by channel comes from GA4 and is read as a trend."
The trap
Spending hours reconciling the two tools down to the euro. They don't measure the same thing, and they'll never agree.
The real risk lies elsewhere: picking, month after month, whichever number suits you. Leadership forgives an explained gap, not a metric that changes definition when it looks bad.
