Measuring the effectiveness of media channels and optimizing the resources invested in them is critical to a firm in obtaining a competitive advantage. While persistence modelling is a well-established approach to this problem, the lack of data often available to marketing practitioners to build these models can reduce their performance. This occurs when a model has too many variables to data points, reducing its generalizability, or when important variables are excluded from the model, reducing its accuracy. Previous research shows that it is important for marketing effectiveness models to capture the complexities of the consumer path to purchase, as well as the direct and indirect effects of media channels. However, it is also important for marketing effectiveness models to be generalizable to new data. We firstly establish a baseline by comparing the differences between persistence models with and without consumer activity data (e.g. Facebook post likes, display advertising clicks), which can help create a more accurate picture of the consumer path to purchase and the relationships between media channels. We subsequently investigate how Bayesian persistence models and networks can improve upon normal persistence models, particularly through the inclusion of an informative prior based on customer information. Overall, we contribute to the literature by measuring marketing effectiveness using a modelling approach that handles practical data limitations, while still allowing for the inclusion of consumer activity data. Additionally, our approach replicates how marketers operate in real-life by starting with a prior set of beliefs about media channel effectiveness, based on customer information, and combining this with historical and ongoing marketing data.