Google Analytics 4: Do I Have to Switch? And If So, How? And When?

Whether you have to switch to Google Analytics 4 is no longer a question. Universal Analytics was switched off in July 2023. If you still have not moved, or you are working with old data, your business intelligence is built on sand.

The question that matters for ecommerce businesses in 2026 is this: when and how do you use GA4 strategically, so its event-based data architecture works for growth? This article explains the fundamental differences between Universal Analytics and GA4, how to build a scalable event model and how to configure ecommerce tracking, including custom events and the BigQuery integration.

Session-based vs. event-based: a new way of thinking

Universal Analytics (UA) thought in sessions and pageviews. A session was a time-limited container that collected hits such as pageviews, events and transactions. The model was intuitive but rigid: it could not combine app and web data natively, it depended on cookies, and it buckled under the complexity of modern customer journeys.

GA4 thinks in events. Every interaction, whether a page view, click, scroll or purchase, is an event with optional parameters. This architecture has four strategic advantages:

• One model across platforms: web, iOS and Android use the same data model. A purchase is a purchase, whatever the device.
• Less dependence on cookies: through modelling, machine learning and Google Signals, GA4 delivers usable data even when tracking is restricted. That is decisive in a post-iOS 14, post-cookie world.
• Flexibility: developers can create up to 500 custom-named events per property, each with up to 25 parameters. That allows a level of detail UA never offered.
• BigQuery export: raw data can be streamed into Google BigQuery. That opens up SQL-based analysis, lifetime attribution and linking to CRM data, features that in UA were only available with paid 360 licences.

The downside: the learning curve is steeper. Reports that took three clicks in UA now mean building explorations or using BigQuery. The training pays off, though: shops that use GA4 systematically reduce their dependence on third-party cookies and build an infrastructure where they own their data.

The event model in detail: structure and naming

GA4 distinguishes four categories of event:

The recommended events are the key for ecommerce. Google has defined a standard naming scheme for retail. Use it and you get preconfigured reports in the Monetization tab and machine learning models trained on those event names. Stray from it, for example “addToBasket” instead of “add_to_cart”, and reports stay empty and attribution models fail.

Each event can carry up to 25 parameters. For the purchase event, these include transaction_id, value, currency, tax, shipping and an items array with product_id, item_name, item_category, quantity and price. That level of detail allows SKU-level analysis straight from the GA4 ecommerce events.

Ecommerce tracking: implementation from the ground up

Ecommerce tracking is usually implemented through Google Tag Manager (GTM). The advantage: marketing teams can adjust events and triggers without a developer deployment. The disadvantage: misconfigurations are common and hard to debug.

The right sequence for a GA4 ecommerce implementation:

1. Build the data layer: the developer implements a JavaScript dataLayer.push on every relevant page. For the product page: event “view_item” with the matching product data.
2. Configure GTM triggers: in GTM, set a trigger on the custom event “view_item”. This trigger fires the GA4 event tag.
3. Map the variables: map data layer variables such as ecommerce.items[0].item_name to GA4 event parameters.
4. Test with GA4 DebugView: before the container goes live, validate every event in GA4’s DebugView. This is where you see whether parameters are missing, wrongly formatted or firing twice.
5. Check the ecommerce reports: after 24 to 48 hours, check the Monetization reports for plausibility. Compare GA4 revenue with your shop backend. Differences of more than 5 % point to implementation errors.

Common errors: the purchase event fires twice because the confirmation page is reloaded. The currency is not in ISO 4217 format (for example “EUR” instead of “€”). The transaction_id is missing or not unique, which causes deduplication problems. You can avoid these if a GTM specialist supports the implementation and validates it against a test plan.

Custom events and conversion tracking

Standard events cover 80 % of needs. For more nuanced analysis, though, custom events are essential. Examples from practice:

• wishlist_add: measures how many users add products to their wishlist without buying. A high wishlist count with low conversion points to price sensitivity or missing trust signals.
• size_guide_open: shows how many customers are unsure about sizing. If 40 % of visitors open the guide but only 10 % buy, the problem lies in the fit description or the size charts.
• loyalty_tier_upgrade: lets you attribute revenue to the loyalty programme and shows which touchpoints lead to upgrades.

Custom events can be marked as conversions directly in GA4. The limit is 30 conversion events per property (50 in GA4 360). Choose those 30 carefully: every event defined as a conversion feeds into attribution and changes how your channels’ performance is judged.

Another advantage of GA4 is its more flexible attribution. UA used last click by default; GA4 uses data-driven attribution. This model evaluates every touchpoint in a journey and distributes the conversion value according to its contribution. For shops with long decision cycles, such as furniture or electronics, this gives more realistic insight than single-touch models.

BigQuery integration: owning your data

Exporting GA4 raw data to BigQuery is the platform’s most powerful feature. In BigQuery you can join GA4 data with CRM data, advertising costs and stock levels. The result is a single data foundation for business intelligence.

A practical use case: a shop exports its GA4 events to BigQuery every day and joins them with Shopify orders via the transaction_id.

The result is a report that shows which channel generated the click and which channel is actually profitable, after returns, payment fees and advertising costs. That kind of net profit attribution is impossible in a standard analytics tool.

Setting it up requires a Google Cloud account and switching on the export in the GA4 property settings. Costs arise only for storage and queries in BigQuery, which for most mid-sized shops come to a single-digit euro amount per month.

Common mistakes when migrating to GA4

Migrating to GA4 is technically demanding. These are the mistakes we see most often in practice:

1. Duplicate purchase events: if the confirmation page is reloaded or reached again via the browser’s back button, the purchase event fires a second time. Fix: a unique transaction_id and server-side deduplication.
2. Missing currency: GA4 expects the “currency” parameter in ISO 4217 format (“EUR”, “USD”). Missing or wrongly formatted values leave revenue reports empty.
3. Non-unique transaction_ids: if different orders get the same ID, for example because a test system and the live system are out of sync, GA4 cannot aggregate the transactions correctly.
4. Ignoring data retention: by default GA4 deletes user-level data after 2 months. For cohort analysis, raise this to 14 months.
5. No testing in DebugView: many teams publish GTM containers without validating events in DebugView. The result is days or weeks of faulty data that cannot be corrected after the fact.

Data protection and Consent Mode: staying compliant in Europe

For German-speaking markets, using GA4 in line with data protection law is central. The European Court of Justice has made clear in several rulings that transferring data to the US is lawful only under strict conditions. Google’s answer is Consent Mode v2 and server-side tagging.

Consent Mode v2 distinguishes between “analytics_storage” and “ad_storage”. If a user does not consent to tracking, GA4 sets no cookies but models the data from server-side signals.

This modelled data is aggregated, but good enough for trend analysis and steering budgets. According to Google, with an opt-out rate of 30 %, modelled data still delivers 70 % of the insight full tracking would provide.

Server-side tagging (server-side GTM) goes a step further: data is not sent from the browser straight to Google but through your own server, usually in the EU.

That reduces the risk of unlawful data transfers and improves data quality, because ad blockers no longer play a part. The downside: running a server takes technical know-how and costs about 50 to 200 euros a month, depending on traffic volume.

A pragmatic timeline for the switch

If you still have leftover UA data or your GA4 implementation is incomplete, we recommend these steps:

• Weeks 1–2: audit your existing GA4 property. Are all recommended ecommerce events implemented? Are custom dimensions and custom metrics configured correctly?
• Weeks 3–4: switch on the BigQuery export and write your first SQL queries for the key KPIs (revenue per channel, cohort retention, product performance).
• Weeks 5–8: move attribution to the data-driven model and compare the results with the old last-click model. Adjust budget allocation based on what you learn.
• Weeks 9–12: train the marketing team in GA4 explorations. The goal: the team can build funnel analyses, path analyses and cohort reports on its own, without relying on outside analysts.

Conclusion

GA4 is not an update to Universal Analytics. It is a new platform with a fundamentally different data model. Switching has long been mandatory; using it strategically is your choice.

The Academy track Google Analytics 4 for CRO ↗ shows you how to put GA4 to work for conversion optimisation. Treat GA4 as a pure reporting tool and you waste its potential.

Use its event-based architecture for custom events, the BigQuery integration and data-driven attribution, and you build an analytics infrastructure that is privacy-compliant, scalable and ready for the future. Investing in a clean implementation and a trained team usually pays for itself within a quarter through sharper budget control and more efficient marketing.