Building a GA-based Customer Propensity Model Unlocking Growth Opportunities for


Incubeta partnered with to leverage Google Analytics (GA) data seamlessly transferred via BigQuery. The objective was to transform this data into actionable insights, specifically identifying audience segments and behavioral features with a high propensity of conversion in the immediate future.


We used built-in BQ ML features to give each website visitor a ‘score’, indicating how likely this visitor was to convert in the next few weeks. From the hundreds of signals that GA provides, we experimented with various feature sets, combined with historical data, to train a deep neural network to generate a model that had a solid predictive power. We then used this scoring to build custom audiences to activate across GMP. 

Keep up to date with our monthly insight digest