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

The problem

The client is a telecoms reseller selling SIM plans under more than one brand. It had a click-through prototype of the sign-up journey that looked finished, and a business that wanted to start selling.

Behind the screens there was no real data, no infrastructure and no tests. Every plan, order and customer came from tens of thousands of lines of made-up data. The app had to handle several branded sign-up journeys and create real orders.

What OES did

Made it real. The mock data came out and real services went in, with production on AWS, built with Terraform, about ten days after work started.

Made a fast launch safe. Before launch OES added a guard so production can never be seeded with test data, wrote a runbook so someone else can operate it, and prepared each brand’s domain in infrastructure code. When the production environment went down during the build, the fix went into the infrastructure code the same day and was written up as an incident.

Kept going after launch. In the first week of real orders, customers weren’t getting confirmation emails. Nothing in the launch checks had looked for them. OES added branded confirmation and operations emails with safe retries, and backfilled the ones that had been missed. Tests, a dependency-security playbook and the architecture checks from an earlier project followed, and a teammate later put the checks in CI on every pull request and in front of production deploys.

The outcome

The app went from prototype to production in about ten days and has taken real customer orders since July, with the operational basics in place: tests, known vulnerabilities cleared, a runbook and infrastructure as code. The one launch gap was found in the first week of orders, fixed and backfilled.

What this means for you

A prototype that looks finished is the riskiest kind, because everyone believes it is nearly done. Before a launch date, check what is still fake, whether someone else could run it tomorrow, and whether anyone has followed a real customer through every email they should receive.