Telecoms sign-up app: prototype to production in under two weeks

The starting point

The prototype looked finished. It had no real data, no infrastructure and no tests. Behind the screens were about 29,000 lines of made-up data.

The client is a telecoms reseller selling SIM plans under more than one brand, so the app had to handle several branded sign-up journeys and create real orders.

Getting to production

I replaced the mock data with real services, set up the production environment on AWS with Terraform, and had it running about ten days after starting. Before launch I added a guard so production could never be seeded with test data, wrote a runbook, and prepared the custom domain in Terraform.

When production went down with server errors, I traced it to the database connection and auth configuration and fixed it the same day. The fix went into the infrastructure code, with an incident write-up so the next person knows what happened.

After launch

The app has taken real customer orders since July. Once orders were coming in I found that customers weren’t getting order confirmations. I built branded confirmation and operations emails with safe retries, and sent the confirmations that had been missed.

Tests went from none to 123. Known dependency vulnerabilities went from nine, five of them high severity, to zero, and I wrote a playbook for keeping them there.

I also ported the architecture checker from another project: 16 rules that stop routes and UI code reaching into the server or database, and require every remote function module to have a helper and a test. A teammate later wired it into CI on every pull request and in front of production deploys. It still passes today.

What a client gets from this

A prototype turned into a product people can buy from, in under two weeks, with tests, security and operations done properly rather than left for later.