Large estates fail migration not because of technology but because of sequencing. The goal of wave planning is to front-load learning while back-loading risk.
We score every database on three axes: business criticality, integration coupling and change volatility. Wave one takes low-criticality, low-coupling systems that still exercise the full toolchain end to end.
Coupling matters more than size. A 200 GB database with fourteen inbound integrations is far harder to move than a 12 TB warehouse with one nightly load.
Reserve the last wave for anything tied to the financial close calendar, and never schedule a cutover inside a quarter-end freeze. The calendar always wins.