Getting started with graph data science experiment design

Hello! First time poster here. I work at a company that uses Neo4j in production. I took many graph theory courses in school, but they were focused on mathematical properties of graphs in the abstract, rather than property graphs for applied use.

I am starting a project investigating how we can use the GDS algorithms on our data. I figure I need to learn about how to design experiments in general, and how to choose projections of our data intelligently so that I can run experiments to select the right algorithm for the job. I've ordered the Barabási network science book, and I am looking for additional resources to skill up in this topic. Thanks!

have you reviewed the neo4j GDS documentation and the free graph algorithms book?

I reviewed some of the former, and I ordered a hard copy of the latter. My main concern is less about the algorithms themselves, and more around being able to provide evidence that any recommendation system we develop is in fact performing favorably compared to a human expert.