Why Cloud Migrations Fail, and What to Do About Each

Common failure modes, and the test for each

• Moving the waste: idle servers and oversized instances are copied as they are, and the bill goes up. *Test early:* measure utilization for at least 30 days before any move.
• The dependency nobody listed: a month-end batch job, a certificate, a DNS record or a partner's IP allowlist breaks at cutover. *Test early:* trace network flows and scheduled jobs across a full business cycle.
• Cost with no owner: spend grows because nobody reviews rightsizing or commitment decisions. *Test early:* name a cost owner and a monthly review before the first wave.
• The one-way cutover: everything moves at once, and the way back was never tested. *Test early:* give each wave a rollback trigger and rehearse the rollback.

How to de-risk, and the questions to answer first

Run discovery that produces a dependency map, not just an inventory. Move in waves that can be reversed. Right-size as you go. Put cost governance in place on day one. A partner who has done this before can turn unknowns into a plan, but the plan still needs your owners to sign it.

Before you set a date, answer these in writing:

• Which dependencies did you find by tracing traffic and jobs, rather than by interviewing people?
• What is the rollback trigger for each wave, and who can call it?
• Who owns cost after cutover, and how often is it reviewed?
• How many minutes of interruption can the business accept, and where is that number written down?

The TiCON safety net, and what we do not promise

A cutover can be designed to limit user-visible impact, and each wave can have a rollback path. Whether a given cutover is visible to users depends on the workload. We do not promise zero downtime. Decide the acceptable interruption, in minutes, before you set the date, and write it where the cutover owner can see it.

TiCON designs cutovers to limit user-visible impact, with a rollback for each wave. Discovery is followed by a fixed quote, and FinOps starts on day one. AWS funding programs may apply to a migration; eligibility is AWS's decision. Engagements are quote-based and billed in USD; other currencies are by contract, so ask.

Related reading

[The 6 Rs of Cloud Migration: How to Assign Each Workload](/resources/blog/6rs-of-cloud-migration) Rehost, replatform, repurchase, refactor, retain and retire, decided workload by workload.

[Cloud Migration Checklist: Ten Gates Before Cutover](/resources/blog/cloud-migration-checklist) The gates to pass before data moves and before cutover, with the evidence each gate needs.

[Reading AWS Pricing: Cost Drivers and a Worked Example in USD](/resources/blog/aws-pricing-bdt-examples) The main cost drivers, and how partner pricing and invoicing change the picture.