Choosing what moves, and how
Every workload gets one of four treatments. Rehost copies it as is, which is fastest and keeps the old design problems. Replatform moves it onto a managed database or a managed runtime, which costs a few extra weeks and removes patching work. Refactor rewrites parts of it for the cloud, which only pays off for systems you will keep for years. Retire switches it off, and in our experience a surprising share of an estate falls into this last group once someone asks who still uses it.
The inventory stage decides the split. We scan the servers, read the connection logs and interview the people who rely on each system. The result is a table with one row per workload and a recommended treatment, an effort range and a risk rating. You approve the table before any money goes on infrastructure.
Most mid-sized estates move in three or four waves over 8 to 16 weeks. A single application with one database can move in 4 to 6 weeks including rehearsals. The order matters more than the speed: shared services such as identity and file storage move first, because everything else depends on them.
