GA4 audit checklist in 25 checks

Tagging and GTM checks verify that the intended collection path runs on each relevant page and only under appropriate conditions. Inventory hardcoded tags, container tags, plugins, and server routes together. A duplicated installation can inflate events even when each individual configuration appears reasonable when inspected on its own.

Check page source and browser requests for Google tags, GTM containers, store plugins, and other collection routes. Fix overlapping installations by choosing an intentional architecture and removing redundant firing paths. Keep required platform tags, but document their destinations and purpose.

Check that expected tags run on landing, product, form, and checkout pages where permitted. Fix overly narrow triggers or missing templates in GTM or site code. Use preview mode and network evidence to confirm the correction beyond the homepage.

Check page_view behavior during client-side navigation, including page_location and page_referrer. Fix overlapping history and enhanced measurement triggers, or missing route handling. Retest direct entry, internal navigation, and browser back actions so each meaningful view follows the intended counting rule.

Check whether the same action reaches GA4 through both browser and server routes. Fix unintended duplication at the sending architecture and verify event destinations. Do not assume arbitrary events deduplicate automatically; purchase transaction IDs address a specific ecommerce case, not every event type.

Event and key event checks confirm that recorded actions match the business outcomes used for optimization. Validate event names, parameters, counting methods, and successful completion conditions. Keep diagnostic interactions distinct from commercial milestones, then verify that reporting and advertising destinations use the intended definition rather than a convenient button click.

Check documented event names, required parameters, and consistent capitalization in DebugView and network requests. Fix typos and inconsistent naming at the source. Register needed custom definitions for reporting and verify that parameter values contain no prohibited personal information.

Check which events are marked as key events and whether their counting method matches the intended outcome. Fix accidental selections and repeated triggers. Record the change date so a different counting configuration is not mistaken for improved conversion performance.

Check that lead events follow confirmed submission rather than a button click, failed validation, or a thank-you page reload. Fix the trigger to use the successful response or reliable application event. Reconcile recorded leads with accepted CRM records and document exclusions.

Check assigned event values and distinguish lead estimates from booked revenue and pipeline. Fix reused default values or inconsistent units. Maintain a value dictionary before sending events to bidding systems, then review the chosen marketing kpis with the commercial owner.

Ecommerce checks verify that purchase and refund events represent actual transactions with usable identifiers, values, and item details. Inspect the data layer and outgoing requests, then compare results with the store ledger. A populated revenue report can still contain duplicated purchases, incorrect currencies, or omissions from alternate checkout paths.

Check that transaction_id is stable, nonempty, and unique to the order, including after refresh or revisiting confirmation. Fix IDs that change between sends or are reused across orders. Verify duplicate behavior with controlled test purchases and the authoritative order record.

Check purchase value and currency against the store definition, with shipping, tax, and discounts treated consistently. Fix string formatting, currency mismatches, and incorrect totals in the data layer. Compare equivalent revenue measures rather than expecting unlike accounting totals to match.

Check item identifiers, quantities, prices, and relevant ecommerce steps across standard and express checkout. Fix missing item arrays or event timing that reads stale cart state. Retest the paths customers actually use, including transitions to hosted or external checkout pages.

Check that refund events reference the correct transaction and affected items where applicable. Fix missing refund delivery and document how cancellations enter each reporting system. Reconcile net revenue over a mature reporting window rather than subtracting refunds from unrelated order periods.

Use reconciled orders as the revenue foundation for roi tracking. Channel reports from multi touch attribution can allocate credit only within their observed outcome boundary. The attribution model guide explains why model changes affect credit even when store revenue stays constant.

Consent mode checks verify that default states, user updates, and tag behavior match the implemented consent design. Inspect actual requests and storage behavior for each choice, rather than relying on the presence of a banner. Technical validation supports implementation review; it does not establish that every legal obligation has been satisfied.

Check that default consent states are established before relevant tags act on them. Fix sequencing in the consent initialization path and align defaults with the reviewed regional implementation. Retest first visits in a clean browser so stored choices cannot hide timing problems.

Check accept, reject, granular settings, and later withdrawal for the applicable storage categories. Fix missing or delayed updates between the consent platform and Google tags. Confirm basic or advanced mode behavior matches the intended design, including requests sent when storage consent is denied.

Check ad_user_data and ad_personalization alongside analytics_storage and ad_storage in the applicable implementation. Fix incomplete mappings from banner choices and inspect downstream tag behavior. Ensure a grant for analytics does not silently become permission for unrelated advertising use.

Advertising link checks verify that analytics outcomes reach the intended accounts with consistent conversion definitions. Review account linking and conversion imports together because the same purchase may enter bidding through different routes. Preserve diagnostic visibility while selecting the intended optimization actions, and confirm that consent choices continue to govern signal use.

Check linked Google Ads accounts, auto-tagging, click identifier preservation, and imported GA4 key events. Fix incorrect destinations or redirects that strip identifiers. Review primary and secondary conversion settings so a native purchase action and an imported action do not inflate bidding totals.

Check Meta Pixel, Conversions API, LinkedIn Insight, TikTok, and Microsoft UET against the channels actually used. Fix missing or duplicate installations and platform-specific deduplication where applicable. Platform attribution totals need not equal GA4; reconcile definitions before labeling a difference as missing revenue.

UTM and channel grouping checks confirm that acquisition labels survive the customer journey and resolve into understandable reporting categories. Standardize naming, preserve relevant identifiers through redirects, and inspect actual landing URLs. Treat channel classification as a documented business rule so naming changes do not silently reshape historical comparisons or hide paid acquisition.

Check external campaign links for consistent utm_source, utm_medium, and utm_campaign values and verify their arrival on the destination. Fix naming drift and redirects that strip parameters. Remove campaign tags from internal navigation so customer journeys are not polluted by self-created acquisition labels.

Check unassigned traffic, paid channel mappings, payment referrals, and custom channel rules against real journeys. Fix malformed source and medium combinations or inappropriate rules. Repair cross-domain and referral handling before exclusions, then compare definitions using the marketing metrics guide.

GA4 data quality checks compare measured events with independent business records and explain differences using consistent definitions. Investigate filters, missing data, processing delays, and reporting limitations before calculating performance. Keep reconciliation separate from attribution, because a channel allocation model cannot repair orders that never entered the eligible outcome data.

Check staff traffic rules, developer traffic, and active filters before excluding records. Fix overly broad conditions and evaluate filters in testing mode before activation. Verify legitimate customer traffic remains included, and document irreversible exclusions that affect future interpretation of reports.

Check distinct transactions and accepted leads against store or CRM records using aligned dates, currencies, and eligibility. Fix missing delivery, repeated triggers, or identifier mapping before comparing acquisition cost. Revisit cost per acquisition advertising after validating the outcome denominator.

Check retention settings, reporting identity, thresholding notices, processing status, and export coverage where configured. Fix unsuitable settings within available options and annotate gaps. Verify that historical coverage supports the cohort or forecast question before using predictive marketing analytics or marketing mix modeling software.

A GA4 reconciliation formula quantifies an observed gap after reporting windows and definitions are aligned. It helps prioritize investigation without declaring every difference a tracking failure. Compare distinct eligible outcomes, separate modeled and observed reporting where possible, and keep supporting order or CRM evidence with the audit log for retesting.

Worked example. Suppose the store records 240 eligible orders and GA4 records 216 distinct matching purchases for the aligned period. Observed order coverage is 216 ÷ 240 × 100 = 90 percent. The unmatched count is 240 minus 216 = 24 orders, or 24 ÷ 240 × 100 = 10 percent.

Inspect those orders by checkout path, device, and consent state using permitted data. Consent choices and blockers can limit observable coverage, so complete equivalence is not an automatic target. Retest repairs against confirmed outcomes instead of assuming a higher event count means better accuracy.

Keep the audit log in marketing dashboards, and use an ai data analyst to investigate supported channel questions. Compare marketing analytics platform plans for ongoing monitoring. DataMarketer checks visible tags, duplicate properties, consent signals, internal UTMs, and tracking coverage across inspected pages; authenticated settings and completed transaction behavior still need the manual review above.