A word that has to earn its keep
“Production-grade” is one of the most abused phrases in software. Every agency claims it; very few define it. For systems that monitor offshore assets or keep a distribution network visible, the definition can’t be vibes — it has to be a checklist someone can audit.
Here is ours.
The checklist
It fails loudly, in the right place. Every failure mode we can anticipate has a defined behaviour: retry, degrade, or alert. Nothing fails silently, and alerts go to a human who can act — not to a channel nobody reads.
It can be observed without SSH. Metrics, logs, and traces are in place before go-live, with dashboards for the two audiences that matter: the operators who use the system and the engineers who maintain it.
It can be deployed on a Tuesday afternoon. Deployments are automated, repeatable, and boring. If shipping a fix requires a maintenance window and a prayer, the system isn’t production-grade — it’s production-fragile.
It survives the departure of its authors. Documentation, architecture diagrams, and a real handoff. The test is simple: can the client’s own engineers make their first meaningful change without calling us?
Why we hold the line
Cutting any of these corners is invisible at launch and expensive six months later — usually at 2am, usually at the worst possible time for the operation. The economics of critical systems reward doing it properly the first time.
That’s also why we put these commitments in writing in every engagement: they are testable, and we expect clients to test them.