Why Your Paid Media Numbers Don't Add Up — And What To Do About It

Why the Numbers Are Different

Each platform uses its own attribution model, its own measurement window, and its own definition of a conversion. None of them are using the same starting point.

• Meta defaults to a 7-day click and 1-day view attribution window. If someone sees your ad without clicking and buys within 24 hours, Meta claims credit. If someone clicks and buys 6 days later, Meta also claims credit. Both of those are legitimate attribution claims — they just overlap heavily with other channels.
• Google Ads uses data-driven attribution by default with a 30-day conversion window. It assigns credit to Google ad interactions that were part of the conversion journey — including touchpoints from weeks before the purchase.
• GA4 uses its own data-driven attribution model and reports independently of any single ad platform. It will distribute credit differently to both Meta and Google Ads.
• Your CRM or ecommerce platform reports actual orders — no attribution model, no modelling, no overlap. Just what happened.

The result: if you sum all the conversions your ad platforms claim, you will almost always get a number higher than your actual order count. This is double-counting across channels, and it's endemic to cross-channel digital advertising. It's not fraud — it's attribution.

The iOS Problem and Why It Made This Worse

From iOS 14.5 onwards, Apple's App Tracking Transparency framework required apps to request permission before tracking users across apps and websites. The majority of users decline. This broke Meta's browser pixel for a significant portion of mobile traffic.

Meta's Conversions API (CAPI) was the primary fix — sending conversion data server-side rather than relying on the browser pixel. But CAPI requires proper implementation, correct deduplication logic between the pixel and the API, and ongoing maintenance. Many accounts are still running pixel-only or have CAPI set up without deduplication, which means:

• Conversions from iOS users are under-reported if pixel-only (ROAS looks worse than it is)
• Conversions are double-counted if CAPI and pixel both fire for the same purchase (ROAS looks better than it is)

Either way, the Meta-reported numbers are unreliable without a correctly configured server-side setup.

What We Actually Do

We approach attribution as a triangulation problem rather than a "which platform is right?" question. No single source tells the full story, but combining multiple signals gets you close enough to make good decisions.

1. Shopify / CRM order data is the truth. Everything else is attribution. We start every analysis here.
2. CAPI with deduplication. We ensure Meta receives the strongest possible signal without double-counting browser pixel and server events. A transaction_id is used as the deduplication key.
3. GA4 with Consent Mode v2. Cross-channel view with modelled data for non-consenting users, giving a more complete picture than pixel-only.
4. Regular incrementality tests. We periodically pause one channel in a geo-controlled test to measure its true incremental contribution to actual revenue. This is the gold standard for understanding what a channel actually delivers.

Incrementality Testing — What It Is and Why It Matters

Incrementality testing is underused and underrated. The premise is straightforward: if you turn off a channel and revenue stays flat, that channel wasn't driving incremental purchases. If revenue drops proportionally, it was.

Running structured tests — geo-based (pause in region A, continue in region B) or time-based holdout tests — gives you real data about channel contribution that no attribution model can replicate. We run these quarterly for clients with sufficient volume. The results consistently produce more actionable insights than optimising within any single platform's own reporting.

Practical Rules for Better Attribution

• Set Meta attribution to 7-day click only. Remove 1-day view from your standard reporting. It inflates numbers significantly.
• In Google Ads, audit your conversion actions. Confirm you're not counting the same purchase twice — e.g., a GA4-imported goal and a separate Google tag conversion for the same event.
• Set a monthly reconciliation: compare total ad platform-reported conversions to actual orders. Track the ratio over time. A sudden change means something in your tracking changed.
• Use blended revenue efficiency as your north star: total revenue divided by total ad spend across all channels. This is a number you can actually trust and act on.

Frequently Asked Questions

Should I trust Meta's reported ROAS?

Use it directionally, not as an absolute number. Meta's reported ROAS will typically overclaim relative to the true incremental impact of your campaigns, particularly for retargeting audiences where customers would have converted anyway. Cross-reference with actual order data and consider running holdout tests for your highest-spend campaigns.

What is the Meta Conversions API (CAPI) and do I need it?

CAPI is Meta's server-side tracking solution. Instead of relying on the browser pixel (which is blocked by iOS ATT, ad blockers, and browser ITP), CAPI sends conversion data from your server directly to Meta. If you're running Meta campaigns, yes — you need it. Without it, you're under-reporting conversions for a significant portion of your audience and your campaign optimisation is working with incomplete data.

What is an incrementality test in paid media?

An incrementality test measures the true additional revenue generated by an ad channel by comparing a group exposed to your ads against a control group that isn't. The most practical version for small-to-medium advertisers is a geo holdout test: pause campaigns in one region while keeping them live in a comparable region, then compare revenue between the two. The revenue difference is the incremental contribution of your campaigns.