Most CMDB schemas start from infrastructure. You discover servers, VMs, network devices, and databases, then try to figure out which application they belong to. The schema mirrors what scanners find, not what delivery teams need.

Product-centric design inverts this. The Product is the root. A server exists because a product needs it. A database exists because a product stores data in it. The taxonomy mirrors the questions teams actually ask: what version is at this site, not what is running on this VM.

This matters when you are managing multiple products at the same customer site. Infrastructure-centric schemas force you to use tags or custom fields to associate servers with products. Product-centric schemas give each product its own branch, so queries are clean and unambiguous.

We built CMDB-Kit around this principle after watching a flat, tagged schema fall apart within a month of managing three products.

Has anyone else tried both approaches? What was your experience?