Every use case listed on this article followed the following nominal pattern: (1) Start with a relational database. (2) Build product until you have a deep un

Every use case listed on this article followed the following nominal pattern: (1) Start with a relational database. (2) Build product until you have a deep understanding of your market and their needs. (3) Move to a specialized data store. For established products, or well-researched use-cases, sure, pick a specialized data store. It will probably serve your needs better. If you don't yet have product-market fit, use a relational database. It will give you far greater flexibility when you discover that "what we thought the market wanted" is different from "what actually made the right numbers go in the right direction".