Low Feature Usage Does Not Mean Safe to Delete, Dependency Graphs Show Why

A software engineering analysis argues that usage metrics alone are unreliable for deciding whether to retire a product feature, because adoption and dependency are fundamentally separate concerns. A real-world B2B SaaS case illustrates this: a "Custom Export Templates" feature used by only 1.4% of accounts was nearly deprecated, but a pre-removal dependency trace revealed it quietly powered a separate feature used by 9% of accounts and an internal billing script tied to enterprise revenue recognition. Two enterprise contracts worth roughly $380,000 in annual recurring revenue also referenced the underlying capability, none of which appeared in any usage dashboard. Rather than eliminating the feature entirely, the team removed only the direct user interface while preserving the underlying engine, saving around 30 engineering hours per quarter instead of risking a costly incident. The analysis concludes that teams must map both technical and business dependencies before sunsetting any feature, not after.