Why your conversion tracking is probably wrong
Six defects account for most broken tracking setups. How to spot each one in under an hour, and what each costs you in wasted spend.
Short answer
Most accounts have at least one of six defects: duplicate pixel and server events, missing consent handling, wrong purchase values, conversions counted more than once per order, missing click IDs, or an optimisation event too far up the funnel. Each one biases the algorithm and inflates cost per acquisition.
Key takeaways
- Tracking defects do not just distort reports — they actively misdirect the bidding algorithm.
- Duplicate events are the most common and most damaging defect.
- Purchase values that include VAT and shipping overstate ROAS by 20–30%.
- Consent-denied traffic must be modelled consistently, not silently dropped.
- An hour of checks usually finds something worth several thousand euros a month.
Defect 1: duplicate pixel and server events
When the browser pixel and the Conversions API both fire without a shared event ID, every purchase counts twice. Reported ROAS doubles, the algorithm learns from phantom conversions, and budget decisions are made on fiction.
Check it in Events Manager: compare the number of purchase events to actual orders for the same day. A ratio near 2.0 is conclusive.
Defect 2: consent handled inconsistently
With Consent Mode v2 configured correctly, denied traffic is modelled and reported consistently. Configured badly, conversions simply vanish for a portion of users, and reported performance drifts with cookie-banner acceptance rates rather than with actual results.
Symptom: a sudden step change in conversions on the day the banner changed, with no matching change in orders.
Defect 3: wrong conversion values
Purchase value should be consistent across every system and should exclude VAT and shipping unless you deliberately choose otherwise everywhere. Mixed conventions inflate ROAS by 20–30% and make platform and store numbers impossible to reconcile.
| Element | Recommended |
|---|---|
| VAT | Excluded |
| Shipping | Excluded |
| Discounts | Deducted |
| Currency | Explicit on every event |
| Refunds | Sent as negative or excluded consistently |
Defects 4–6: the quieter ones
Fourth, conversions firing on page load rather than on order confirmation, so a refreshed thank-you page counts twice. Fifth, click IDs not captured on the landing page, breaking offline conversion feedback later. Sixth, optimising toward an event too far up the funnel — add-to-cart or page view — which trains the algorithm to find browsers rather than buyers.
- Fire purchase events on server-confirmed orders, not on page render
- Capture GCLID, FBCLID, WBRAID and UTM values on entry and persist them
- Optimise toward purchase or qualified lead once volume allows
- Verify event match quality in Events Manager monthly
A one-hour audit checklist
Run these checks in order. Most accounts fail at least two, and each failure has a direct euro cost in misallocated budget.
| Check | Typical impact if failing |
|---|---|
| Platform purchases vs real orders | Up to 100% overstated ROAS |
| Event ID present on both pixel and CAPI | Duplicate optimisation signal |
| Consent Mode v2 signals on every page | Unstable, drifting reporting |
| Value excludes VAT and shipping | 20–30% overstated ROAS |
| Click IDs stored with each lead | No offline feedback possible |
| Optimisation event is bottom-funnel | Higher CPA, weaker audiences |
Fix the signal before anything else
This is why the first stage of our Signal Loop method is measurement, not media. Optimising creative and budget on top of a broken signal simply finds the wrong people more efficiently.
Our free ad account audit runs exactly these checks and comes back within 48–72 hours of getting access, with a prioritised list of what to fix first.
FAQ
How can I tell quickly whether events are duplicated?
Compare a single day's platform purchase count with the actual order count in your store back-end. A ratio close to 2.0 for Meta almost always means the pixel and Conversions API are firing without a shared event ID.
Should conversion value include VAT?
Usually not. Excluding VAT and shipping gives a value closer to the revenue you actually keep and makes ROAS comparable to your break-even calculation. The critical part is applying the same convention everywhere.
Does fixing tracking improve performance or just reporting?
Both. Reported numbers become trustworthy, and because bidding algorithms learn from these events, cleaner signal usually reduces real cost per acquisition within a few weeks.
Keep reading
Want this run on your own account?
Request a free ad account audit. We review tracking, account structure, creative mix and wasted spend, and send back a prioritised 90-day plan within 48–72 hours of getting access.