The problem this solves
Every marketing decision downstream of analytics inherits the tracking's flaws, and most teams have never audited the tracking itself. GA4 says what it says, dashboards repeat it, and budgets move on numbers nobody has cross-checked since the setup was configured, often by someone who has since left, under a consent framework that has changed twice. The data feels authoritative because it is precise, and precision is exactly what broken tracking fakes best.
The failure modes are quiet by nature. Events that stopped firing after a site update leave no error, only absence. Double-counted conversions inflate a channel until someone reconciles against a source of truth, which is a task without an owner. Consent-management holes silently drop whole segments of visitors from measurement, skewing every ratio computed afterward. None of these announce themselves; they surface, if ever, as a discrepancy someone notices in a meeting and nobody has time to chase.
The audit that fixes this is a reconciliation: GA4's account of reality checked against independent witnesses - Search Console's view of what search actually delivered, and server-side signals of what the site actually served. Where the accounts disagree, there is a finding; where the finding has a cause, there is a fix; and each fix lands as a task with severity attached, because a tracking bug's importance depends entirely on which decisions it is quietly poisoning.
How the mission runs
- Inventory the measurement setup. The audit starts with what GA4 is configured to believe: the events, conversions, and channel definitions in place, and what each is supposed to capture. This inventory is the checklist the reconciliation runs against, and configuration that no longer matches the live site gets its first flags here.
- Reconcile against Search Console. Search Console's record of search-delivered traffic is compared against GA4's account of the same visitors. The two measure differently by design, so the audit checks whether the divergence matches the expected pattern; a gap beyond the explainable range points at dropped tracking, consent holes, or misattributed channels.
- Check internal consistency and double-counting. GA4 is checked against itself: conversions that fire multiple times per real action, events double-counted through tag duplication, and totals that disagree between reports that should agree. Each anomaly is verified by reproduction where possible, so the finding describes a mechanism rather than a suspicion.
- Probe the consent and coverage holes. The audit examines what the setup structurally cannot see: traffic segments dropped by consent configuration, pages missing tags entirely, and platforms or regions where measurement quietly stops. Coverage holes are sized where the evidence allows, because a hole's severity is its share of the picture.
- File findings as severity-ranked tasks. Each verified finding becomes an Asana task: the discrepancy, the evidence, the likely mechanism, the proposed fix, and a severity grade tied to which reported numbers it distorts. The audit lands as a work queue ordered by how much each bug is costing decision quality.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A set of Asana tasks, one per verified tracking finding, each stating the discrepancy, its evidence, the likely mechanism, the proposed fix, and a severity grade tied to the decisions the bug distorts. Behind the tasks, a reconciliation summary showing where GA4, Search Console, and server-side signals agree and where they diverge beyond explanation, so the team knows which dashboard numbers to trust while the fixes land.
Make it yours
- Re-run the reconciliation after the high-severity fixes ship, confirming the divergences actually closed and catching any fix that moved a number the wrong way.
- Schedule it quarterly as a standing sanity check, since site releases and consent changes reintroduce tracking drift on their own schedule.
- Scope a focused run to conversion events only ahead of a budget cycle, when the numbers about to justify spending deserve the strictest check.
Frequently asked questions
What counts as a real finding versus normal measurement noise?
The audit expects divergence and knows the explainable patterns: GA4 and Search Console measure differently by design, and consent frameworks legitimately reduce coverage. A finding is a divergence beyond those expected ranges with a reproducible mechanism behind it, and each task states why the gap is beyond explanation.
Does the audit change my analytics configuration?
No. It reads, reconciles, and files tasks; every fix is implemented by your team through your own change process. Tracking configuration is production infrastructure, and the audit treats it that way: findings arrive with proposed fixes and evidence, and the change decision stays human.
How is severity decided?
By decision impact rather than technical drama. A subtle double-count on the conversion event that justifies budget outranks a loud error on an event nobody reads. Each task names the reports and decisions its bug distorts, which is what lets a lead order the fix queue with confidence.