Two studios moved their delivery and billing onto newer platforms this year, drawn by lower unit pricing and a dashboard that promised to consolidate monitoring. Both moved back within twelve months, and neither cited price as the reason.

The first moved because a recommended caching configuration behaved differently under a live event, and the incident took four hours because the runbook described the old system. The second moved because a billing migration produced duplicate charges for 300 subscribers, and the rollback took a week.

What boring actually means

Fewer moving parts, a runbook that matches reality, dashboards people already know how to read, and a vendor whose support line answers at 3am. It does not mean old. It means predictable, documented and reversible.

“We did not save money. We bought a system nobody on the team could debug at 3am.”Operations lead who reversed a migration

The measurable version: incidents per quarter, mean time to recovery, and the share of incidents resolved without escalating to the vendor. All three studios that stayed on their older stack reported better recovery times than the two that migrated, largely because the on-call engineer knew where the logs were.

The practical test before a migration is not a benchmark. It is a question: can the person on call at 3am fix this without reading documentation they have never seen? If the answer is no, the savings are theoretical and the risk is not.

Key takeaways

  • Two migrations were reversed within a year for operational reasons, not cost.
  • Recovery time, not unit price, is the metric that decides whether a platform stays.
  • The test: can the 3am on-call fix it without new documentation?
TL

Tobias Lund · Technology editor

Hosting, delivery and moderation systems. Former platform engineer, still reads the status page first.