Our quality dashboards are green. Coverage is up and lint is clean. Last week a small pricing change touched five services and needed four teams to agree before we could ship it. The deploy still felt dangerous. We measure file-level quality well. We do not measure architecture at all. Gartner put architecture and dependencies in scope for its first Magic Quadrant for Technical Debt Management Tools. We can compute change coupling from version control history. Some teams track blast radius or services touched per change. What metric actually made your coupling visible enough to act on?
We do not measure architecture at all.
Services touched per change made coupling visible for us. We logged it for every production change and small fixes averaged 3 services for months. A tianpan.co piece on hidden dependency edges named it: It’s drawing the graph. Our checkout flow proved it. I still have no clean way to count shared database edits in that metric.
three services never moved anyone. counted teams needed to agree instead. got decoupling funded.
We counted teams needed to agree. Invites shrank immediately. The number dropped fast. Coupling stayed brutal. What metric actually survived that gaming?