Logistics & Supply Chain
Northwind Freight
From one fragile server to boring, measurable infrastructure
Northwind's dispatch platform — the system 40 drivers and 3 warehouses depend on — ran on a single VPS that one contractor knew how to deploy. We moved it to AWS with infrastructure-as-code, CI/CD, and observability, without a minute of scheduled downtime.
The whole company ran on a server nobody dared to touch
Northwind's dispatch and tracking platform lived on one oversized VPS. Deploys were a 45-minute manual ritual performed by a single contractor, over SSH, usually on Sunday nights. There was no staging environment, backups were a cron job nobody had ever restored from, and the monthly hosting bill kept climbing because the only scaling lever was "buy a bigger box."
When that contractor gave notice, what had been an inconvenience became a business risk with a date on it.
Two weeks of mapping before we changed a single thing
We started the way every engagement starts: a written audit. We mapped the deploy ritual step by step, traced which services actually talked to each other, load-tested the database, and — critically — restored a backup to prove whether the safety net existed. It half did: the database backup was sound, but uploaded proof-of-delivery documents had never been backed up at all.
The audit gave Northwind a prioritized risk list they could act on with or without us. They chose with us.
Everything became code before anything moved
Every piece of the new AWS environment — VPC, ECS services, RDS Postgres, S3 document storage, IAM roles — was written as Terraform before it was created. Secrets moved from a shared text file into AWS Secrets Manager. The application was containerized so that "works on the contractor's laptop" stopped being a deployment dependency.
From that point on, staging and production were provably identical, rebuilt from a Git repository instead of from memory.
Deploys went from a Sunday-night ritual to a merge button
GitHub Actions now runs the test suite on every pull request, builds a container image, deploys it to staging automatically, and promotes to production on approval — with one-click rollback to any previous image. The 45-minute manual deploy became a 4-minute pipeline anyone on the team can run, and Sunday nights went back to being Sunday nights.
Prometheus and Grafana watch the fleet, with alerts tuned to page for real problems — failed dispatch jobs, database saturation — rather than noise.
Cut over in reversible steps, verified with real traffic
We refused the big-bang cutover. The new environment ran in shadow for two weeks, mirroring production traffic while we compared outputs. DNS moved one subdomain at a time, database replication kept both environments in sync, and every step had a written, tested rollback. Drivers and dispatchers never saw a maintenance page.
Right-sizing during the move — smaller instances, reserved pricing, S3 lifecycle rules on old documents — cut the monthly bill 38% even after adding staging and monitoring that had never existed before.
Infrastructure their own team can run — and prove
Northwind's platform now deploys in 4 minutes, has held 99.95% uptime since cutover, and survived its first real incident — a bad migration — with a 90-second rollback instead of a lost evening. Documents are backed up and restore-tested quarterly, automatically.
We delivered runbooks, architecture diagrams, and two training sessions with their new in-house developer, then moved to a light monthly monitoring arrangement. They own every repository, every AWS account, and every credential.
Built With
The audit alone paid for itself — it found a backup gap that would have ended us. Six months later our own developer runs everything, exactly like they promised.Operations Director · Northwind Freight
More Case Studies
Want Results Like These?
Tell us what you're building — or what's breaking. We'll come back with a concrete plan and a fixed price.
Book a Free Consultation© Copyrights 2023-2026 | Axomble | All Rights Reserved
