#Bolf #GA4 #consentmode #datamodeling
Thanks to modeling, we recovered a large portion of data on traffic and transactions in the online store.
The history of Denley dates back to 2007, and the first online store under the Denley.pl brand was established in 2010 (14 years ago!). Its enormous success in the country spurred international expansion. Thus, in 2011, the first e-store in Germany under the Bolf brand was launched. So far, the brand has opened 17 independent e-stores across Europe!
Challenge:
On March 6, the Digital Marketing Act came into effect, requiring data collection only from users who consent to it. This posed a significant risk that much less data essential for marketing analysis and website optimization would be sent from online stores to Google tools. To counter this, we implemented a Consent Management Platform for the client and configured advanced Consent Mode for Google tools to enable data modeling.
We are proud of:
Recovering data through data modeling:
- Approximately 31% of sessions
- Approximately 23% of purchases
Our Actions:
We implemented advanced consent mode on the client’s website, allowing Google systems to continue collecting anonymous signals about site traffic even when cookies are declined. Data collected in this manner is automatically sent to Google servers, where, once a sufficient amount is gathered, it is modeled and then incorporated into Google Analytics and Google Ads reports.
According to Google documentation, the threshold conditions required for modeling to start are as follows:
- The GA4 service records at least 1000 events daily with the parameter analytics_storage=’denied’ for at least 7 days.
- The GA4 service has at least 1000 daily users sending events with the parameter analytics_storage=’granted’ for at least 7 out of the last 28 days.
In the case of Bolf’s GA4 service, where we later conducted the analysis, the monthly site traffic was sufficiently high to give a good probability of automatic modeling activation.
A few weeks after implementation, we verified the reports presenting the number of sessions and purchases, comparing the data displayed in mixed identity for reporting purposes (with modeling) and in observed categories identity (without modeling). This allowed us to capture the amount of data modeled and subsequently recorded in Google Analytics reports.
- The increase in data from most main traffic sources observed after enabling modeling is approximately 27% to 37% more reported users and sessions. Smaller traffic sources recorded a higher percentage increase in data.
- The increase in data from most main traffic sources observed after enabling modeling is approximately 28% to 60% more reported transactions and revenues. Smaller traffic sources recorded a higher percentage increase in data.
Effect:
Data verification showed that modeling in the described store operates efficiently, mitigating the potential lack of data from declined user consents.
- Thanks to consent mode, purchase reporting in GA4 is approximately 23% more accurate.
- Thanks to consent mode, session reporting in GA4 is approximately 31% more accurate.
It is worth noting that modeling can “return” different amounts of data to reports. Depending on the amount and quality of input data, modeling may, after about a month, load modeled data into reports that, in total, give almost 100% coverage of traffic or much less. It can vary in each case depending on the industry and the volume of recorded traffic.