A data-driven approach to Google Advertising Campaigns

Digital advertising increasingly relies on data-driven decision-making, and companies show a growing need for tools capable of translating performance metrics into more targeted and efficient budget strategies. This thesis addresses this need by developing a predictive framework, implemented in a web-based prototype, to support budget allocation in Google Ads campaigns using daily performance data provided by the company Krick.com. The study integrates exploratory data analysis, time-series forecasting models and multivariate methods to identify the factors that influence campaign performance. Prophet and Random Forest models show that the Click-Through Rate can be reliably predicted at the channel level (MAE ≈ 0.004–0.005). These forecasts enable Synapse, an interactive prototype that generates personalized budget recommendations based on sector, region and campaign objectives. The multivariate analysis based on Principal Component Analysis and clustering reveals that campaigns do not form distinct groups but instead lie along continuous gradients of scale and efficiency, supporting the decision to model performance at the channel level. Overall, the findings demonstrate that predictive analytics can significantly enhance budget allocation decisions by providing interpretable and forward-looking insights. The Synapse prototype offers a practical foundation for more adaptive and data-driven marketing planning.