Before you optimize ads using conversion data, verify that your tracking can answer one basic question:
For a real order, can you explain what each system received and why the numbers differ?
Use this checklist before blaming attribution, installing another pixel or rebuilding your tracking stack.
1. Pick the business source of truth
For ecommerce, your order platform/database should usually be the baseline for:
- completed orders;
- order ID;
- gross/net revenue definition;
- currency;
- refunds/cancellations.
Ad platforms and analytics systems are not accounting systems.
2. Define the exact conversion
Write down what counts as a Purchase.
Examples:
- checkout completed;
- payment authorized;
- payment captured;
- order created;
- cash-on-delivery order accepted.
If teams use different definitions, dashboard numbers will never reconcile cleanly.
3. Check the event exists
For a sample of real orders:
- did a Purchase event fire?
- was it browser, server or both?
- which tool sent it?
Missing event = collection/delivery problem.
4. Check for duplicates
For every order:
- count browser Purchase events;
- count server Purchase events;
- inspect event IDs;
- test reload/retry behavior.
One order should not silently become several logical Purchase events.
5. Check order/transaction ID
Verify that:
- it is present;
- it is stable;
- it matches the store;
- it is not reused incorrectly.
6. Check value
Make sure Purchase value follows one defined business rule.
Ask:
- include tax?
- include shipping?
- include discount?
- gross or net?
- before or after refund?
7. Check currency
Wrong/missing currency can make revenue comparisons meaningless.
8. Check products/items
Where relevant:
- content/product ID;
- quantity;
- item value;
- category;
- SKU/variant consistency.
9. Check customer identifiers
Only use accurate, permitted customer information.
Measure coverage rather than checking one successful test event.
10. Check click identifiers
For paid traffic, verify relevant click identifiers survive:
- landing;
- session;
- checkout;
- payment redirect;
- final Purchase.
11. Check browser identifiers
If a destination relies on browser context, confirm the identifiers are captured and forwarded where appropriate.
12. Check consent
Document:
- what happens before consent;
- after consent;
- when denied;
- when consent state changes.
Server-side tracking does not remove consent obligations.
13. Check each destination separately
Do not assume an event delivered to one platform was delivered to all.
For each:
- received?
- accepted?
- rejected?
- delayed?
- missing fields?
14. Separate raw events from attributed conversions
Meta or Google Ads can attribute a conversion differently from the store’s order count.
First validate raw event delivery.
Then investigate attribution.
15. Check timezone
A midnight boundary can create apparent daily mismatches.
Record timezone for:
- store;
- analytics;
- ad platform;
- reporting tool.
16. Check refunds/cancellations
Decide whether reporting should:
- send refund events;
- adjust revenue;
- keep original conversions;
- use another business reporting layer.
17. Check payment methods
Test different paths:
- card;
- wallet;
- redirect payment;
- cash on delivery if relevant;
- accelerated checkout.
A “works on my test card” result is not enough.
18. Check mobile and desktop
The same checkout can behave differently across devices.
19. Check overlapping apps/plugins
List every tool that can send:
- Meta Purchase;
- Google Ads conversion;
- GA4 purchase;
- other destination events.
Disable or scope overlapping ownership intentionally.
20. Create a regression process
Re-test after:
- theme changes;
- app/plugin updates;
- checkout changes;
- consent-platform changes;
- GTM changes;
- destination API updates.
How to classify discrepancies
Use these categories:
Missing event — order exists, event absent. Duplicate event — one order creates multiple logical events. Value mismatch — event exists but revenue/currency differs. Identity loss — event arrives with weak/missing identifiers. Attribution difference — raw event exists, platform credits it differently. Reporting definition — dashboards calculate totals differently.
Do not call every discrepancy “attribution.”
A simple audit sample
Pick 10–20 orders across:
- mobile/desktop;
- different traffic sources;
- payment methods;
- new/returning customers where permitted.
Trace each order end to end.
Small, high-quality order tracing often reveals more than comparing two dashboard totals.
Where KePixel fits
The Tracking Health Audit should operationalize this checklist before recommending a larger tracking implementation.
Current source and destination availability should be confirmed before any platform-specific implementation recommendation.
Recommended next step
If you cannot map which systems send Purchase today, start there before installing another app or CAPI integration.
