Rethinking Attribution After GA4

Google Analytics 4 (GA4) represents the biggest update to Google's analytics platform in years, replacing Universal Analytics. GA4 introduces an entirely new event-centric data model and interface focused on understanding user behavior.

While GA4 may have a better underlying data model, its user experience feels rushed and incomplete, causing marketers to rethink their attribution stack. Beloved reports and key functionality are missing, and basic data visualization rules have not been followed. As a sign of how well the migration is going, they recently implemented this obnoxious countdown timer:

UX issues aside, this change is set against a rapidly evolving attribution landscape, in which the role of analytics in general is diminishing. With GDPR and the rollout of iOS14, there will be irreparable holes in your tracking no matter how powerful your analytics platform. As marketers are forced to rethink analytics, they’re also exploring alternate methods like marketing mix modeling and conversion lift testing.

This post will explore how GA4 differs from Universal Analytics, challenges with its rollout, and how a mixed approach can provide the best insights into marketing ROI. The future of attribution is a combination of tools and methods, not reliance on any single platform.

How GA4 Differs from Universal Analytics

GA4 features an event-centric data model focused on how users interact with websites and apps rather than pageviews. Universal Analytics had events and user IDs, but GA4 is built around them, allowing for native cross-device tracking and a unified view of the customer journey across multiple channels. The updated interface offers a new reporting suite and a free integration with BigQuery for querying raw data. Overall, GA4 provides a broader and more detailed look at user behaviors and journeys.

However, this new approach also means abandoning the familiar pageview-centric model of Universal Analytics and learning an entirely new system. All implementation, customization, reporting, and analysis will need to be rebuilt in GA4. For marketers used to Universal Analytics, this massive change will require significant time and effort to adopt, and there are at least 40 outstanding issues to be addressed. What’s more, it feels like basic principles of data visualization were not followed. Tell me for example what school of UX design teaches you to use 5 near-identical shades of blue / purple in a time series chart?

GA4 also aims to address some previous concerns with how Universal Analytics collected and stored data. By assigning users a unique ID not associated with personally identifiable information, GA4 says it reduces the risk of compromising user privacy while still allowing cross-device tracking. However, some marketers and privacy experts argue anonymized IDs also pose risks and that GA4 still collects a concerning amount of data. There are several ongoing legal and political battles ongoing, including several in Europe where Google Analytics has been found in violation of GDPR.

Challenges with GA4's Rollout and Adoption

GA4's launch has lacked functionality compared to the mature Universal Analytics, which may reduce initial usage and slow long-term adoption, especially among smaller companies with fewer technical resources. While the setup process has been streamlined, GA4 requires reimplementing tracking, reconfiguring views and reports, and relearning how to analyze data. For casual or inexperienced GA users, this steep learning curve could discourage adoption.

Some companies may also hesitate to transition to GA4 due to uncertainty around its data collection and privacy policies following changes like GDPR and privacy laws in the US. Even though GA4 claims to be compliant with all laws and address previous concerns, its approach to assigning unique IDs and tying behavior across devices could still deter privacy-conscious organizations or individuals. If key users opt out of data collection due to privacy concerns, attribution models become less accurate.

Approximately 60% of the top 100,000 websites have Universal Analytics installed, but competitors in the analytics space must be kicking themselves for their good fortune. On the higher end platforms like Matomo, Mixpanel, and Amplitude must be seeing huge adoption as a retaliation for forced adoption of GA4, and tools like Segment which make it easy to switch between analytics providers will likely have seen a bump in new customers. On the low end of the scale, there are a number of small privacy-centric startups like Plausible Analytics and Simple Analytics that have done well.

While GA4 may eventually match or surpass Universal Analytics in functionality as Google continues improving the platform, the initial rollout has lacked in some areas. The updated interface and features are appealing, but for now, some marketers may get more value from other tools. GA4 may still make sense as a default or as an additional source for collecting logged-in user data where you have a user ID, but should likely be combined with other methods for non-logged in traffic and overall marketing attribution.

Is Traditional Analytics Even Relevant Anymore?

Even in the ‘good old days’ of pageview-based web analytics, we knew tracking was imperfect. Just because I tracked someone clicking on an ad and buying, does not mean it was the ad alone that caused that sale. Perhaps that person heard about the product from their friend, or had bought before using a different account on a different device. Conversely we know that ads work even when people don’t click on them, as with TV, Radio, and Billboards.

Perhaps the whole conversation is moot, because we’re losing the ability to even see online activity, now that users are increasingly opting out of tracking entirely by using ad blockers, VPNs, or clicking ‘no’ on cookie banners. On mobile the deprecation of the IDFA, a unique identifier for Apple devices, has left a huge hole in the ability to track the most valuable segment of mobile usage.

With AI tools now able to browse the internet, the web traffic recorded in GA4 may one day become mostly bot traffic. If AI replaces search engines, marketer’s attention will shift from driving website traffic or app installs, to increasing the chances of being recommended by ChatGPT, a new discipline I’m calling Artificial Intelligence Optimization (AIO). Just as page views became an unsuitable model with the rise of mobile, perhaps events will need to be retired as AI radically alters user behavior.

A Mixed Approach to Marketing Attribution

While MTA can still be useful for gaining insights into logged-in users, MMM and lift studies do not require the same level of personal data to be effective. MMM uses aggregate data to determine how marketing channels influence revenue and conversions. Lift studies run by the ad platforms Meta and Google still require user-level tracking, but not geo tests, which measure the impact of campaigns by comparing locations where ads ran to control locations where they did not.

These methods can then be combined with MTA for "triangulation"—using them together helps determine where MTA may be inaccurate and calibrate overall attribution models. Lif tests are the gold standard of proving the value of advertising, and can be used to calibrate your marketing mix models, which in turn can tell you how over or under inflated the numbers are you’re seeing in analytics.

For example:

• The results of your lift test indicate Meta Ads has an incremental CPA of $30.
• You work on your MMM until it agrees with the results of your lift test, for that period.
• The model shows Google is overcounting by 20%, so adjust the CPA target to $37.5.

Rather than relying entirely on GA4, marketers should take an integrated approach that incorporates other techniques like marketing mix modeling (MMM) and lift studies. Multi-touch attribution (MTA) alone provides an incomplete picture, especially as privacy regulations and changes make persistent cross-device tracking more difficult. You need to compare your MTA results to MMM and conversion lift studies.

An integrated approach balancing privacy and functionality will become increasingly critical as data regulations and policies continue to evolve. Marketers should not rely entirely on any single tool but rather incorporate multiple methods for the best insights.

Conclusion

GA4 represents a step forward for understanding how users interact with brands, but marketers will need to combine its capabilities with other techniques to determine how marketing dollars are best spent. GA4 enables more accurate cross-device attribution but currently lacks some functionality. A mixed approach leveraging MMM, geo tests, MTA, and a focus on logged-in users may be most effective for marketers seeking the best attribution insights and ROI.

To adapt to changes like GA4 and increasing privacy concerns, marketers should incorporate multiple modeling techniques rather than relying on any single tool. With the right balanced and data-driven approach, marketers can gain a competitive advantage in today's complex attribution landscape. But they must be willing to pair new solutions with tried-and-true techniques and not assume any single platform will meet all their needs.

GA4 is only one part of the larger puzzle that is marketing attribution in an age of privacy and cross-device journeys. But combined with the right supporting methods, its capabilities can help brands better understand and optimize how they reach and influence customers. The future of attribution is integrated, and marketers willing to take an open-minded, mixed approach will be poised to succeed.