CLOUD REPATRIATION TRIAGE & EXECUTION CHECKLIST (2026)

Use this in a 60–90 minute working session with engineering, finance, and security.

1) DEFINE THE DECISION UNIT (DON’T SAY “LEAVE AWS”)
- List 5–10 distinct workloads/services (not vendors): API tier, background workers, ETL jobs, CI runners, observability pipeline, object storage + egress, primary DB, analytics warehouse, internal tools.
- For each workload, write a one-line SLO expectation (availability/latency) in plain language.

2) COST REALITY CHECK (NO HAND-WAVING)
- Identify the top cost drivers per workload (compute, storage, managed DB, network/egress, logging/metrics ingest, support plans).
- Write down which costs are usage-based vs contract-based (commitments/reservations).
- Note any “costs” that are really org design problems (no owner, no deletion, no tagging, everyone can provision).

3) PORTABILITY SCORE (QUALITATIVE IS FINE)
For each workload, mark:
- Stateless vs stateful
- Clear API boundary vs coupled to internal services
- Uses proprietary managed services heavily (yes/no)
- Data gravity risk (large datasets; frequent cross-service transfers)
- Rollback path exists (yes/no)

4) PICK THE FIRST MOVE (LOW BLAST RADIUS)
Choose ONE pilot that is:
- Observable (you can measure errors/latency)
- Easy rollback (traffic shift, feature flag, blue/green)
- Not your primary database
Good candidates: batch jobs, stateless workers, CI runners, non-prod environments, log retention/cold storage policy changes.

5) GOVERNANCE BEFORE MIGRATION (OR YOU’LL RECREATE THE BILL)
- Enforce tagging and ownership (who pays, who deletes).
- Restrict provision permissions; require review for expensive services.
- Set lifecycle rules (expiration for test envs, storage tiers, log retention).
- Define what proprietary services are allowed going forward and the required exit plan.

6) EXECUTION PLAN (ONE PAGE)
- Owner (single accountable person)
- Success criteria (what must be true in prod)
- Rollback plan (specific steps)
- Timeline with milestones (design, staging, cutover, validation)
- Security review items (secrets, network, access, audit evidence)

7) CONTRACT STRATEGY (PARALLEL TRACK)
- Map current commitments to actual steady-state usage.
- Decide what you’ll commit to vs keep on-demand.
- Use credible alternatives (colo quotes, bare metal providers, second cloud) to negotiate—without pretending you’ll move everything.

8) POST-MOVE “NEVER AGAIN” RULES
After the pilot:
- Write 3–5 rules that would have prevented the original cost pain (examples: no unmanaged data egress paths; OpenTelemetry required; time-bound dev environments; managed DB only with explicit owner/runbook).
- Schedule a quarterly workload review: what moved, what got cheaper, what got riskier.

Outcome you want: a repeatable capability to shift workloads and renegotiate from a position of strength—not a one-time migration hero story.