A trader starts checkout for an evaluation, gets approval from the payment provider and closes the tab before the confirmation page loads. The order system still knows the purchase succeeded. A Meta Pixel that fires only on the confirmation page may miss it. Sending the confirmed order through Meta Conversions API gives that sale a server path. If the browser also reports the same purchase, both copies need the same event ID.
Give the browser and the order system different jobs
Google Tag Manager has a web container that runs in the browser and a server container that processes incoming requests. Stape can host the server container. The web container can send eligible checkout activity to a tagging endpoint under your domain. The server container uses a client to interpret the request, applies your rules and sends approved fields to Meta through a CAPI tag. A subdomain such as metrics.example.com shares the site's parent domain. A path on the exact website origin, such as www.example.com/metrics, is a separate deployment option. Google's custom domain guide · Stape's domain setup
Use the browser for actions it can observe, such as starting checkout. Use your backend to confirm financial outcomes. When a payment reaches the accepted state in your order system, the backend can send a minimal event to the server container through an authenticated integration. Where payment and account activation happen separately, keep both timestamps. Finance can reconcile when the evaluation was sold, and operations can see when the trader received access. PropFirmsTech's overview of platform, payment and CRM handoffs
Confirm the purchase in the order ledger
Start with the payment provider's signed webhook or equivalent authenticated status update. Verify it, match it to the intended order and check the amount, currency and final payment state. For a paid Stripe Checkout order, inspect the session's payment status. A delayed payment can finish checkout before funds are available. Wait for the successful asynchronous outcome and the accepted order transition. With another provider, document the equivalent state and confirm how authorization, capture and delayed payments work. A thank-you URL or an initial authorization alone is insufficient evidence of a paid order. Stripe Checkout payment events · Checkout Session fields · Stripe webhook verification
Use the actual paid value under your documented value convention and the order's currency. Specify how discounts, tax and currency conversion should appear in reports. Label sandbox data and keep it out of production purchase totals. If finance records a partial or full refund, retain the original purchase and record the reversal against it. Use GA4's refund event if you measure refunds in GA4. Track net revenue from the ledger, where you can also see pending refunds and disputes. Stripe refund states · GA4 purchase and refund events
Deduplicate in the order system and in Meta
The same webhook can arrive several times. Make the paid-order transition and creation of its measurement record idempotent. Log handled provider delivery IDs, and enforce one purchase record per accepted order in your own data store. Queue delivery after the order is committed. Keep the same measurement record, ID and original event time on each retry. This also protects against separate webhook event types reporting the same payment. Stripe's guidance on duplicate webhook deliveries
Meta handles the second layer. If you report the same purchase via both browser Pixel and CAPI, use the same dataset, event name Purchase and opaque ID for the same order. The Pixel argument is eventID. The CAPI field is event_id. Issue the ID once in your backend and provide it with the verified purchase state when the browser can report that purchase. A reload should never generate a new ID. A second paid order needs its own ID. Meta documents a 48-hour window for matching browser and server copies sent with the same ID and event name to the same Pixel. Your order system remains responsible for its own delivery retries and accounting. Meta's deduplication documentation · Meta's server event parameters
Apply data choices before tags fire
Carry the applicable consent state into the server decision. For a purchase emitted by the backend, keep a reliable record of which measurement uses the visitor granted, and account for withdrawals. Apply your consent rules to Meta separately from Google's consent settings. A tagging endpoint on your domain still needs to respect those decisions. Confirm the business, dataset and intended event are eligible under Meta’s Business Tools Terms, and apply any restrictions you see in Events Manager. Checkout and the transaction ledger need to work when optional measurement is denied. Google's server-side consent guide
Allow only the fields each destination needs. Keep card details, identity-verification documents, account balances, trading history and incidental personal data out of the event and URL. If eligible matching fields are permitted for the use, follow Meta's normalization and hashing rules. Restrict access to logs and keep API credentials server-side. Review preview output and failure logging as well as the final payload. Meta's customer information parameters
Follow one test order through both tools
In Google Tag Manager, open Preview for the Stape-hosted server container. Start a sandbox checkout from the connected browser. Inspect the incoming request, the client that claimed it, the mapped event, the tag decision and the outgoing request. For a backend-originated event, use the active server preview header on the test request or Stape's sGTM Preview Header config. Where available, scope the Stape setting to the test path or IP and restore the original setting afterward. Keep the preview token out of public examples. Google's server debugger · Stape's Preview Header config
Point both your sandbox Pixel and server tag at a dedicated Meta test dataset. Keep that dataset out of live campaigns and audience rules. Open Test Events for the test dataset and include its test_event_code on test sends. Meta can use events carrying this code for ads measurement and targeting, so use the separate dataset to keep synthetic purchases out of production. Compare event name, source, value, currency and ID with the sandbox ledger. Check the browser and server pair and review deduplication evidence. A tag that fires in Preview is one step. Check receipt in Meta separately, then remove the test code and verify that only live traffic uses the production dataset. Meta's Test Events instructions
| Test | Expected result |
|---|---|
| Paid order with return page, then reload | One accepted ledger order. Browser and server use the same Purchase ID. |
| Paid order with no return visit | Eligible server purchase still arrives with the correct value and currency. |
| Decline, abandonment or delayed payment | No purchase until the payment reaches the accepted state. |
| Replayed webhook or tag retry | The existing ledger and measurement records are reused. |
| Optional tracking denied or withdrawn | The order works and the disallowed tags stay off. |
| Partial or full refund | The ledger shows the reversal against the original order. No new purchase fires. |
Before using purchase events for bidding, have engineering and finance reconcile the test orders, then compare a bounded production sample of accepted orders with eligible Meta deliveries. Record consent exclusions, failures and any difference in currency or value. Compare delivery to the event record and keep Ads Manager's attributed results separate. This verifies the handoff without treating a test result as paid campaign performance.
For the broader order-to-GA4 workflow, use the prop firm conversion tracking guide. For help scoping your current stack, see Adtrend's conversion tracking service.
For the operational messages after checkout, see the prop firm email lifecycle guide.