For an online shop owner: when GA4 orders do not match the shop, test one order before you rebuild anything

Describe the failed result in one order

Place one test order, or recall one real order you can find in both systems, and note what GA4 shows for it. Three situations cover most shops. GA4 shows one order under two or more different transaction IDs, or DebugView shows two purchase events for it. GA4 shows revenue with no items, or no funnel before the purchase. GA4 shows nothing at all. Each points to a different bounded repair, and the test is the same: one order, one event, with the right ID, items and value.

You do not need to read code or send customer data to do this. Orders from your own test are the evidence.

• Write down the number of events one order produced.
• Keep GA4 and shop totals for one week, but treat them as secondary evidence.

Who owns what

You own the shop, the GA4 property and the right to publish. Someone must be able to grant a scoped role on GA4 and Tag Manager, and to approve any theme change. The accountable person decides which orders count as test orders and how a test order will be cancelled.

Platforms differ in how much of checkout can be changed; a hosted checkout with no way to send events limits what any adviser can measure.

• Name the person who can approve theme or tracking-setting changes.
• Decide the test product and the test payment route beforehand.

Pick the smallest repair that matches

Doubled purchases fit Fix duplicate GA4 purchases at £295. A missing or partial funnel fits Shop funnel events from £795. If several different things are wrong, the rebuild project may be better; a single key event not recording fits the £195 repair. A scheduled test with a monthly record is £195 a month, on a cadence agreed in writing. All prices are untested proposals, payment follows sign-off, and nothing promises revenue or ad results.

• Start with the one-order test; it tells you which row applies.
• Send the platform name and the totals for one week, not order data.