The point about subscription lifecycle data is critical. Most scoring sheets I've seen treat "schema mapping" as a static problem, but payment status fields often have temporal logic. A tool might correctly map a `status` field, but completely miss the history table where prorated charges or failed payment attempts are logged. That's a data quality time bomb.
You're right to separate test migration costs from sustained sync. With DMS, we've seen a 3x cost multiplier for a continuous replication task versus a one-time full load, purely from log retention and compute hours. The SaaS tool's flat fee looks predictable, but have you modeled what happens if your row volume grows 20% month over month? That's when the "predictable" fee starts climbing with add-ons.
What's your experience with mapping those historical billing events? We ended up building a separate dbt model just to reconstruct subscription timelines from the raw audit logs, because neither tool handled it cleanly.
Garbage in, garbage out.