Introduction
Precise evaluation of marketing measures is increasingly challenging for companies. Are traditional evaluation models like last-click or data-driven attribution models now outdated? Are older approaches like tracking-independent MMM (Marketing Mix Modeling) being brought back to market through advanced ML models? Can server-side tracking help close the growing data gaps?
Or does the best-practice approach lie in a hybrid model that enables optimal advertising budget allocation?
Long-term experience in this field shows what it means when cookie consent suddenly blocks essential data. The challenges are also well-known when detailed tracking approaches can no longer be fully implemented.
The Starting Point: Data-Driven Attribution (DDA)
For several years now, providers like Google have offered the ability to analyze customer journeys using UTM parameters and other website data. Statistical methods are used to create data-driven models for evaluating the performance of individual campaigns.
Challenges of Cookie-Based Attribution
These models are primarily based on cookies that are stored in users' browsers and subsequently analyzed. However, both users' increased privacy awareness and stricter requirements from the EU GDPR lead to significant limitations in data collection.
Only about a quarter of users consent to full tracking (Source).
Modern Approaches to Conversion Attribution
Modern evaluation methods therefore no longer rely exclusively on first-party cookies. Instead, missing information is extrapolated using statistical models. Practical experience shows that combining Google Analytics 4 (GA4), internal systems, and BigQuery can achieve conversion attribution of up to 70%.
Modern tracking models are based on:
- First-party cookies
- Statistical modeling in GA4
- User-Pseudo-ID from GA4, enriched through BigQuery integration
Additionally, raw data from BigQuery can be used to determine repurchase probabilities and linked to internal CRM systems through transaction data (e.g., order numbers) – for a holistic view of customers, orders, and product performance.
Marketing Mix Modeling (MMM)
An alternative evaluation approach is Marketing Mix Modeling, e.g., with Google Meridian (Source). The open-source tool is particularly well-suited for privacy-compliant attribution – completely without user tracking, cookies, or cross-device IDs.
MMM Data Foundation
Instead, MMM uses:
- Aggregated media spending
- Seasonal influences
- Revenue figures
The model allows evaluation of traditional channels like TV or print as well.
Important Parameters in MMM Setup
Assumptions (Priors): For example: "Google Ads influences revenue by 10–30%" – these hypotheses are tested in the model.
Decay Factors: Determine when a campaign reaches its maximum impact (e.g., 2–3 days after TV broadcast).
Saturation Effects: Analyze at what point additional advertising spend brings minimal additional benefit.
💡 Important Note: An MMM model like Meridian can be very meaningful at an aggregated level – provided the input data is of high quality (e.g., clicks, costs, impressions per measure). Modeling at too granular a level poses risks like overfitting or multicollinearity.
Therefore, measures should ideally be aggregated at a higher level (e.g., Google Ads, Meta, TV, promotions).
Hybrid Models as the Optimal Path
While attribution models provide detailed insights, they can also become imprecise when campaign density is too high. In such cases, they closely resemble the last-click model.
The Optimal Combination
Conclusion: The ultimate discipline is a hybrid approach – server-side tracking combined with GA4 for operational decisions and an MMM model (e.g., Meridian) for strategic decisions at an aggregated level.
Successfully Building an Evaluation Model
1. Establish Tracking Foundation
→ Integration of a Tag Manager (Tag & Consent Manager)
2. Data Architecture & Infrastructure
→ For scalable, centralized data processing
3. Interfaces to Marketing Providers
→ e.g., Google, Meta, Criteo, etc.
4. Standardized Naming Convention
→ For consistent data structuring
Technology Stack for Modern Attribution
- Attribution Tools: Google Analytics 4, Adobe Analytics
- MMM Solutions: Google Meridian, Meta Robyn
- Data Processing: BigQuery, Snowflake, dbt
- Orchestration: Apache Airflow, Prefect
Conclusion
The future of marketing attribution lies not in a single model, but in the intelligent combination of different approaches. Hybrid models enable both operational and strategic decisions to be made on a solid data basis – even in times of increasing privacy regulations.