GA4 Configuration Audit Checklist

GA4 property configuration decisions made at setup become invisible six months later — when data retention cuts off, when Google Signals thresholding hides segments in reports, when custom dimensions were never registered for parameters you have been collecting for a year.

This checklist is a systematic review of every GA4 configuration decision. Work through it when you inherit a property, before major reporting periods (annual planning, quarterly reviews), or when your GA4 data produces unexpected results. The whole checklist takes under 90 minutes.

Note

This checklist covers GA4 property configuration specifically. For a broader review that includes your GTM container, consent implementation, data quality, and performance, use the GTM & GA4 Audit Checklist.

---

Section 1: Data Retention and Basic Settings

These are the settings people configure once and forget, which is exactly when they become problems.

Data Retention

• Data retention is set to 14 months

  • Location: Admin → Data Settings → Data Retention → Event data retention
  • 2 months (the default) cuts off data for any Exploration report with a date range beyond two months
  • 14 months is the maximum available on all properties (including free tier)
  • This setting does not affect standard reports — only Explorations

• Reset user data on new activity is toggled based on your use case

  • If enabled: the retention window resets each time a user is active (extends data for active users)
  • If disabled: the clock starts from first event regardless of subsequent activity
  • For most implementations: leave enabled (the default)

Timezone and Currency

• Property timezone matches the primary business timezone (not UTC by default)
  • Timezone affects day boundaries in reports — a mismatch causes sessions to split at the wrong time
  • Cannot be changed retroactively; changing it only affects future data
• Reporting currency matches the currency of primary revenue transactions
  • GA4 converts multi-currency purchases to the reporting currency using daily exchange rates

Industry Category

• Industry category is set accurately
  • Affects benchmark comparisons in Insights and some predictive modeling
  • Low priority but worth confirming it is not set to the default “Other”

---

Section 2: Google Signals

Google Signals enables cross-device user counting and demographic reporting — but it comes with important trade-offs.

• Google Signals activation decision is intentional (not left at default)

  • Location: Admin → Data Settings → Data Collection → Google Signals data collection
  • Enabling: provides cross-device user counts, age/gender demographics, and remarketing audiences that work without cookies
  • Trade-off: activating Signals introduces thresholding — segments with fewer than a minimum number of users are suppressed in reports to prevent individual identification
  • If Signals is enabled and some of your audience segments show “Thresholding applied” warnings, this is why

• Thresholding implications are understood by the team

  • If your properties have high traffic, thresholding rarely matters
  • If you have small user segments you rely on for analysis, Signals thresholding will suppress those segments
  • Mitigation: use BigQuery for granular analysis — BigQuery export is not affected by thresholding

• User Data Collection Acknowledgment is completed (required to activate Signals)

---

Section 3: Reporting Identity

Reporting identity controls how GA4 determines whether two visits are from the same user.

• Reporting identity is configured intentionally

  • Location: Admin → Reporting Identity
  • Blended (recommended for most): Uses User ID first, then Google Signals, then device-based
  • Observed: Uses only signals GA4 can observe directly — User ID and Google Signals — no modeling
  • Device-based: Traditional cookie-based identity, no cross-device matching
  • Most properties should use Blended if User ID is implemented or if Signals is enabled

• User ID is implemented if your application has authenticated users

  • User ID lets GA4 associate sessions from the same logged-in user across devices and browsers
  • Must be set via user_id parameter in GA4 (either in the config tag or via set_user_properties)
  • Critical: Never send PII (email, name, phone) as the User ID — send a hashed or internal identifier only

---

Section 4: Data Streams

Each GA4 property can have multiple data streams (web, iOS, Android). Each needs its own review.

For each data stream:

• Stream is active and receiving data (check Realtime → no gaps in the last hour)
• No duplicate streams exist (two streams for the same website doubles every event count)
• Stream name accurately identifies the source (not “Web Stream 1”)
• Measurement ID is implemented correctly on the corresponding website or app

Enhanced Measurement (web streams)

• Enhanced Measurement settings have been reviewed deliberately — not left on defaults
  • Location: Data Streams → [stream] → Enhanced Measurement
  • Page Views: Leave enabled (this is the base behavior)
  • Scrolls: Review the 90% threshold — is this meaningful for your content?
  • Outbound Clicks: Disable if you have custom click tracking to avoid double-counting
  • Site Search: Enable and configure the correct query parameter (q, s, query, etc.)
  • Video Engagement: Disable if you have custom video tracking; enable for YouTube embeds
  • File Downloads: Appropriate file extensions listed?

---

Section 5: Data Filters

Data filters exclude or modify events before they are processed. Mistakes here permanently affect your data — they cannot be retroactively corrected.

• Internal traffic filter is active (not just defined — check that status is “Active”, not “Testing”)

  • Location: Admin → Data Filters → Internal Traffic
  • Testing mode lets you preview the filter’s impact in DebugView without applying it to real data
  • Leave in Testing mode until you have verified it is working correctly, then switch to Active
  • Common mistake: leaving the filter in Testing mode indefinitely — it does nothing in Testing mode

• The internal traffic IP range is complete

  • Covers office IP, VPN exit nodes, automated testing servers, development machines
  • Check with your IT/DevOps team for the full list of IPs that should be excluded

• Developer traffic filter is configured (for properties where developers use the production site)

  • Uses debug_mode: true parameter to identify traffic
  • Less reliable than IP-based filtering — use both

• No active filters are suppressing legitimate user traffic

  • An incorrect hostname filter, for example, can silently drop a large portion of traffic
  • Validate by checking event counts before and after filter application in a test window

---

Section 6: Custom Dimensions and Metrics

Every parameter you want to use in reports must be registered. Unregistered parameters are collected but not reportable.

• All event parameters used in reporting are registered as Custom Dimensions

  • Location: Admin → Custom Definitions → Custom Dimensions
  • Check your dataLayer specification: every parameter name there should have a corresponding custom dimension
  • GA4 limits: 50 event-scoped, 25 user-scoped custom dimensions per property

• Custom Dimension scope is correct for each parameter

  • Event-scoped: data that changes per event (page_type, product_id, button_text)
  • User-scoped: data that is relatively stable per user (subscription_tier, user_role, customer_segment)
  • If user properties are registered as event-scoped, they work in event-level analysis but not in user-level audience building

• No custom dimensions are registered for parameters that are never sent

  • Wasted allocation — you only get 50 event-scoped dimensions
  • Check by filtering GA4 Explorations by the custom dimension: if it always returns “(not set)”, the parameter is not being sent

• Custom metric registration for numeric parameters (if applicable)

  • Numeric parameters you want to aggregate (sum, average) should be registered as Custom Metrics, not Dimensions
  • Examples: article_word_count, video_watch_time_seconds, form_completion_time

---

Section 7: Key Events (Conversions)

• All business-critical events are marked as key events

  • Location: Admin → Events → toggle “Mark as key event” for relevant events
  • Every event you want to appear in conversion-focused reports needs to be marked

• Counting method is correct for each key event

  • “Once per event”: counts every occurrence (appropriate for purchases, form submissions)
  • “Once per session”: counts once regardless of how many times the event fires in a session (appropriate for session engagement, article reads)

• No automatic events are accidentally marked as key events

  • session_start, first_visit, user_engagement as key events will produce meaningless conversion rates
  • Check each marked key event and ask: “Does marking this as a conversion add business value?”

• Key event counts are being validated against source-of-truth data

  • Monthly: compare GA4 purchase counts to your order management system
  • If GA4 shows 20% fewer purchases than your database, there is a tracking gap to investigate
  • Document the expected discrepancy (returns, test orders, etc.) so the team knows what variance is normal

---

Section 8: BigQuery Export

Free for all GA4 properties. If you are not exporting to BigQuery, you are leaving the most powerful analysis capability on the table.

• BigQuery is linked

  • Location: Admin → Product Links → BigQuery Links
  • Requires a Google Cloud project with the BigQuery API enabled

• Daily export is active (check that export is “Active” not “Paused”)

• Streaming export is considered for near-real-time use cases

  • Streaming incurs BigQuery costs (the daily export is free beyond storage)
  • Most properties do not need streaming — daily is sufficient for reporting

• Intraday tables are being used for same-day analysis (if needed)

  • events_intraday_YYYYMMDD tables are available throughout the day
  • They are replaced by the final daily table the next day

• BigQuery dataset region matches your data residency requirements

  • Cannot be changed after creation

---

Section 9: Google Ads Link

• Google Ads is linked (if you run Google Ads)

  • Location: Admin → Product Links → Google Ads Links
  • Enables: importing GA4 key events as Google Ads conversions, creating audiences in GA4 for Ads targeting, and seeing GA4 metrics in Ads reports

• Conversion import is configured in Google Ads (if using GA4 key events as Ads conversions)

  • In Google Ads: Tools → Measurement → Conversions → Import from Google Analytics
  • Verify that the correct key events are imported and set to the right conversion value

• Auto-tagging is enabled in Google Ads (essential for correct attribution)

  • Auto-tagging appends the gclid parameter to Ads URLs, enabling GA4 to attribute sessions to the correct campaign
  • Without auto-tagging, Google Ads traffic may appear as direct traffic in GA4

---

Section 10: Search Console Link

• Search Console is linked (if you have organic search traffic worth analyzing)

  • Location: Admin → Product Links → Search Console Links
  • Requires Search Console property ownership
  • Enables: the “Organic Search Traffic” report in GA4, landing page performance with organic search metrics

• The correct Search Console property is linked (domain property vs. URL prefix property)

---

Section 11: Attribution Settings

Attribution settings determine how GA4 assigns credit for conversions across the user journey.

• Attribution model is reviewed and appropriate for your business

  • Location: Admin → Attribution Settings
  • Data-driven (default for properties with sufficient conversions): uses ML to assign fractional credit across touchpoints. Requires 50+ conversions per month and 1,000+ events to use
  • Last click: 100% credit to last campaign before conversion (like UA’s default)
  • Google paid channels last click: last click credit only for Google paid channels; organic/direct uses data-driven
  • Most properties with sufficient data should use data-driven. If your property is new or low-traffic, use last click

• Lookback windows match your business’s decision cycle

  • Acquisition events lookback: 30 days (default). Increase if your users research for longer before converting
  • Engagement events lookback: 90 days (default). The window for cross-channel credit
  • Paid channels lookback: 30 days (default for non-YouTube), 60 days for YouTube

---

Section 12: Audiences

• Predictive audiences are active (if your property qualifies)

  • Location: Admin → Audiences → Predictive
  • Available for properties with sufficient purchase data (minimum signals required by Google)
  • Predictive audiences: “Likely 7-day purchasers”, “Likely 7-day churning users” — high value for remarketing

• No audiences are based on PII (email, phone, name in event parameters)

• Existing audiences are still relevant — delete stale audiences from past campaigns

---

Audit Score Summary

Section	Status	Priority Issues	Notes
Data Retention & Settings
Google Signals
Reporting Identity
Data Streams
Data Filters
Custom Dimensions
Key Events
BigQuery Export
Google Ads Link
Search Console Link
Attribution Settings
Audiences

Critical fixes (affects data accuracy today):

• Data retention set to 2 months — change immediately
• Internal traffic filter in Testing mode — switch to Active
• Key event misconfigured (wrong counting method, wrong event marked)

High priority (affects analysis quality):

• Missing custom dimension registrations
• BigQuery not linked
• Attribution model on default when data-driven is available

Medium priority (affects reporting completeness):

• Google Signals not considered for cross-device reporting
• Search Console not linked
• Predictive audiences not activated

---

Related Resources

• GTM & GA4 Audit Checklist — Broader audit including GTM container, consent, and performance
• Migration: UA to GA4 — If this property was migrated from UA, verify migration-specific settings
• Tracking Documentation Template — Template 3 (GA4 Property Configuration Document) helps document what this audit finds