This question is about finding options to model the data for a biologics drug manufacturing. The manufacturing process is divided into upstream and downstream, where the output material from an upstream batch is the input for a downstream batch. To make this complex, upstream materials from a single batch are collected into multiple bags and these bags from multiple batches gets randomly pooled to form a downstream batch. Because of this random intermediate pooling of upstream materials, there is no one to one relation between upstream batch and downstream batches. Now, the problem we are trying to solve is, to identify the root causes of variation in drug product quantity at the end of our downstream process. There are about total 15 individual process steps in upstream and downstream each. The variation in drug product quantity can come anywhere from upstream, downstream, any raw materials, any process step in between. Right now, the data is prepared as 1 downstream batch in a row with weighted averages of upstream batches and materials in individual columns as variables. However, this is not the best way because we are losing some fidelity within upstream batches in the form of weighted averages. I am looking to get alternative ideas from experts of this platform. Below is a pic to demonstrate the genealogy explained here.