Consent Mode v2 and the Google Ads vs GA4 purchase gap
A European outdoor apparel retailer runs Google Ads and GA4 across five country domains. After deploying a Consent Mode v2 banner in March, the team sees Google Ads report 1,240 purchase conversions for April while GA4 reports 1,060 transactions from Google Ads. Enhanced conversions for web is enabled, and purchase is a primary conversion action in Google Ads with a 30-day click window; GA4 uses data-driven attribution and a 30-day click plus 7-day view window. Add-to-cart is also imported as a secondary conversion. The discrepancy grows on mobile. The team wants to keep Smart Bidding stable but needs one source of truth for reporting. They ask whether to pause enhanced conversions, change attribution models, or accept modeling. Budget is €80k/month; purchase value varies by country. The analytics lead suspects consent-related data loss but cannot explain why Ads is higher.
1. What are the most likely causes of the Google Ads vs GA4 purchase discrepancy, and how would you validate each cause before changing configuration?
Read the worked answer
Separate measurement artifacts from real performance. 1) Consent Mode v2: Google Ads can report modeled conversions when users decline ad_storage or analytics_storage, while GA4 may report observed data only. A higher Ads count is therefore expected and is not proof of overcounting. Validate by segmenting consent state and comparing conversion time to click time. 2) Attribution: the Google Ads purchase action may use last click and a 30-day click window, while GA4 uses data-driven attribution and a 30-day click plus 7-day view window. Compare the same action and window using the Google Ads conversion report and GA4 Advertising > Attribution. 3) Counting: Google Ads can count every conversion; GA4 counts transactions. Check the action's count setting and whether add-to-cart is included in the Conversions column. 4) Enhanced conversions: normally improves match and can add conversions; check diagnostics and deduplication rather than pausing it. 5) Time zone, currency, and cross-domain measurement. Validate with a BigQuery export, GCLID-level reconciliation, and a short consent-accepted test. Align on one conversion definition for bidding and a reconciled definition for cross-channel reporting.
2. The team proposes switching all Google Ads conversion actions to data-driven attribution to match GA4. What trade-offs should they weigh, and what would you recommend?
Read the worked answer
Switching is not a like-for-like match with GA4 and can reset learning. Google Ads data-driven attribution uses its own interaction data and modeled conversions; GA4 uses GA4 session data and its own models. Smart Bidding optimizes the selected conversion action's attribution, so a change alters the bidding signal and may cause a temporary performance dip. View-through and engaged-view treatment can also differ. Recommendation: keep a stable primary purchase action with a consistent model for bidding, especially during peak season. Use data-driven attribution if path complexity is high, but document the definitional gap. For reporting, export GA4 and Google Ads data to BigQuery and reconcile at transaction level. Use add-to-cart only as a secondary observation action, not a bidding action. If leadership needs one number, report platform orders and explain the modeled-conversion bridge.
3. What guardrails would you put in place before accepting consent-mode modeled conversions as the source of truth?
Read the worked answer
Monitor consent rate by country, modeled vs observed conversion ratio, and week-over-week variance. Set an alert if Google Ads conversions exceed GA4 by more than 25% for two consecutive weeks, or if the gap changes abruptly after a banner update. Use a blended truth metric such as ecommerce-platform orders and reconcile monthly. Keep the primary purchase action for bidding, but do not optimize to modeled conversions alone without validation. Segment mobile and desktop because consent behavior differs. Document that Google Ads conversions may include modeled data while GA4 reporting may not. Run a geo holdout only if the gap changes a major budget decision; otherwise, track the reconciliation monthly and avoid changing models mid-quarter.