The situation
World Sailing — the world governing body for the sport of sailing — ran its entire digital estate inside a long-standing supplier’s cloud account: the main website, a member portal serving nearly 139,000 accounts, administrative systems, public data APIs and the pipelines that ingest live results from Olympic timing providers. The organisation depended on infrastructure it did not own, could not inspect and could not change. When the supplier relationship came to an end, everything had to move: infrastructure, source code, data, deployment pipelines, identity and DNS, without interrupting services used daily by sailors, member national authorities and event organisers.
Why the existing approach was limiting
Little of the estate existed in a form World Sailing controlled. Infrastructure existed only as live resources in the supplier’s account, with no infrastructure-as-code and limited documentation. Credentials were embedded in application configuration. There was no web application firewall. Operational knowledge sat with the supplier, and every change depended on them. A simple lift-and-shift would have reproduced that uncertainty in a new location; the migration was the moment to replace it with an estate the organisation could own, inspect and govern.
Signal North’s role
We designed and delivered the migration end to end with a small senior team. Because the estate was largely undocumented, the first task was reconstruction: we reverse-engineered the running system and raised 37 formal information requests with the outgoing supplier — source code, deployment pipelines, container images, database snapshots, configuration — tracking every one to closure. The estate was then rebuilt from scratch as around 16,000 lines of Terraform: networks, databases, container platforms, content delivery, storage, DNS and security — more than 300 resources in all. Every recovered configuration value was verified against the live production system rather than assumed.
Key decisions and intervention
- rebuild rather than lift-and-shift, so the new estate was fully defined as version-controlled code from day one;
- run a structured information-request process with the outgoing supplier, converting undocumented knowledge into verified configuration;
- build make-before-break: the complete new estate ran and was smoke-tested on internal domains before any production DNS changed;
- design the DNS so rollback required no third party — production names pointed at records World Sailing controlled;
- adopt the existing identity tenant rather than migrate users, preserving 138,949 accounts with no password resets;
- harden in transit: a web application firewall where none existed, IAM roles replacing embedded credentials, encryption at rest and least-privilege access throughout;
- cut over ten public websites and services in a single night, inside a two-hour content freeze, against a 63-step runsheet with a formal go/no-go gate and a seven-day rollback window.
What changed
World Sailing now owns its estate outright: its own AWS account, its own source repositories and deployment pipelines, its own identity contract, and a security posture materially stronger than the one it left behind. The infrastructure is defined in code, documented and reviewable, so any engineer — internal or external — can understand and change it safely. The supplier’s environment was decommissioned a week after cutover, on schedule. Five weeks after the first line of infrastructure code was written, the organisation was running on foundations it controlled.