Why GA4 and Meta never agree

Two rulers, two totals#

Meta answers which ads earned attention and assisted outcomes across people, devices and views; GA4 answers which sessions and last clicks preceded site behaviour. One is person-centric and credit-sharing, the other session-centric and winner-takes-all on the last click. Expecting identical totals mistakes two instruments for one broken one. Start at the Meta Ads hub for the family view.

Stop chasing parity and start translating: learn each ruler, then reconcile both against the order book that actually banks. Teams that accept two honest lenses argue less and decide faster. Meta own attribution windows guide shows Meta side of the lens.

Clicks versus sessions#

One ad click can spawn several sessions across devices and days, or zero sessions when the app opens instead of the browser. Meta counts the click and its downstream outcome; the analytics platform counts sessions that arrived with recognisable campaign tags. Tag gaps, redirects that strip parameters, and in-app browsers each widen the click-to-session gap silently.

Audit the tag chain end to end: every ad URL tagged distinctly, redirects preserving parameters, landing pages firing the analytics tag on every template. Fix the plumbing before debating philosophy; half of all mismatches we review are broken tags, not deep methodology. Pixel setup keeps Meta side clean.

Views that GA4 never sees#

Meta default reporting blends view-credited conversions that session-based last-click reporting never grants: someone sees the ad, converts days later via direct visit, and Meta claims a view-through while the analytics platform credits direct. Neither is lying; the credit rules simply differ. Split Meta click and view columns to quantify the view share before any cross-platform comparison.

Video-heavy prospecting carries the largest view share, so compare its Meta totals against click-only columns first. If click-only still overshoots analytics substantially, the remaining gap lives in windows, privacy or tags. Attribution windows set the credit rules.

Privacy gaps on both sides#

Consent refusals, blocker usage and platform restrictions remove events from both systems asymmetrically: Meta fills some gaps with modelled conversions and AEM aggregation, while the analytics platform samples, thresholds or drops its own share. Each system estimates differently, so privacy-heavy segments diverge most. Developer context on Meta side is at AEM developer docs.

Track the divergence itself: platform-to-truth ratios per channel, monthly, show whether gaps grow or hold steady. Growing gaps name consent or capture decay; stable gaps are methodology you can plan around. Modelled conversions explains Meta estimates.

The reconciliation routine#

Monthly, in one sitting: pull Meta click-only conversions per campaign, analytics last-click conversions per matched campaign tag, and order-book revenue with cancellations removed. Compute each platform-to-truth ratio, note window and model settings beside each, and file the sheet where next month will find it. One hour, same template, every month.

Investigate only meaningful moves: sudden ratio swings name tag breakage, redirect changes or consent shifts, while slow drifts name methodology and privacy trends. Confirm clean dedup on Meta side before blaming the other platform. Server cover tightens Meta side of the ratio.

Reporting both without war#

Give each platform its job in reporting: Meta for creative and audience reads inside fixed windows, analytics for site behaviour and last-click journey analysis, the order book for revenue truth, and lift tests for causality. No single number runs the business; the triangulation does. Stakeholders presented with labelled lenses stop demanding impossible parity.

Set targets per lens rather than translating one target across systems: a Meta ROAS target, an analytics CPA guardrail, and a blended business target on booked revenue. Conversion Lift settles the arguments totals cannot, and our team builds the routine when internal debate stalls.

Frequently asked questions#