We relied on the interview-driven methodology defined in our previous work \cite{benvenuti2022}, which allowed us to specify various simulation scenarios to frame the boundaries of all possible pipeline executions. Then, we generated this simulated event logs in the traditional XES format using the Simio (https://www.simio.com/), obtaining 10,000 execution traces compliant with the simulation scenarios. In the picture it is shown the Directly-Follow Graph (DFG) representing the pipeline structure, discovered by feeding a process discovery tool (https://fluxicon.com/disco/) with the simulated event log. The pipeline is triggered when the system receives a request to model a new marketing campaign or to report on how an already existing one is performing. In both cases, the first two steps of the pipeline are querying the required data and to apply specific transformations on it. Then, if the request was for a report there is the need of merging the queried data, while for the request of a model an algorithm to compute it is launched. Next, if the request was for a model, the result needs to be stored. Finally, the pipeline ends with either the report or the model being generated.