A true story of good intentions, missing architecture, and a backend that almost took the whole system down. A true story of good intentions, missing architecture, and a backend that almost took the whole system down. A true story of good intentions, missing architecture, and a backend that almost took the whole system down. The Setup The Setup We were brought in to audit an internal logistics platform that had been under development for five years. The budget? Roughly €15,000 per month — totaling just under €900,000. At first glance, everything seemed functional. But once we started digging into the code and infrastructure, we found a system on the edge of collapse. Not because of bad intentions — but because of years of well-meaning decisions made under pressure. What Went Wrong What Went Wrong The backend was a PHP monolith on a single server — no scalability, no failover.
The frontend was built on Next.js, but with no structure, routing logic, or user flow. Users couldn’t find anything.
The mobile app? Built by someone without mobile experience. No search, no filters, no UX — just chaos.
The data sync relied on CSV files exported from Oracle 9, transferred via FTP, and re-imported daily.
All logs were dumped into unpartitioned Oracle tables. Eventually, they grew too large to query — and rebuilding indexes meant shutting down the database. In production. The backend was a PHP monolith on a single server — no scalability, no failover. backend was a PHP monolith The frontend was built on Next.js, but with no structure, routing logic, or user flow. Users couldn’t find anything. frontend was built on Next.js The mobile app? Built by someone without mobile experience. No search, no filters, no UX — just chaos. mobile app The data sync relied on CSV files exported from Oracle 9, transferred via FTP, and re-imported daily. data sync All logs were dumped into unpartitioned Oracle tables. Eventually, they grew too large to query — and rebuilding indexes meant shutting down the database. In production. This wasn’t just technical debt. It was digital quicksand. Why This Happens Why This Happens We see this pattern often: MVPs that were never upgraded
Legacy systems duct-taped together
Internal tools that cost hundreds of thousands, but don’t actually solve problems MVPs that were never upgraded Legacy systems duct-taped together Internal tools that cost hundreds of thousands, but don’t actually solve problems And always the same request: “Can you build us an MVP in 3 months?” Yes — but only if we agree what minimum and viable really mean. minimum viable What It Took to Fix It What It Took to Fix It Here’s what we proposed after the audit: Upgrade to Oracle 11 and partition the bloated tables
Offload logs to an external system (ELK or ClickHouse)
Rebuild the backend as a microservice architecture (Spring Boot, Docker, K8s)
Implement real CI/CD pipelines
Design a service API layer for all interactions
Map out user journeys and business processes from scratch
Secure the infra with encrypted VPNs, IAM, and zero-trust access Upgrade to Oracle 11 and partition the bloated tables Offload logs to an external system (ELK or ClickHouse) Rebuild the backend as a microservice architecture (Spring Boot, Docker, K8s) Implement real CI/CD pipelines Design a service API layer for all interactions Map out user journeys and business processes from scratch Secure the infra with encrypted VPNs, IAM, and zero-trust access Not a quick fix. Not a cheap one. But a path forward that could actually scale. We’re still working with this client today — not just rebuilding software, but helping them rethink how they operate digitally. The Takeaway Tech debt isn’t always the result of bad developers. It’s often the outcome of survival mode — building just enough to stay afloat. But eventually, systems that were never designed to scale… don’t. You don’t always have to rebuild. But if you do — do it right.

How €900,000 Disappeared Into an App No One Really Used (And What We Learned Fixing It)

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

How to Write a Full-Stack Application to a Backend Engineer

Why Would Anyone Call a Race Condition Nice?

The In‑Depth, Crystal‑Clear Roadmap to Learning System Design End‑to‑End (HLD & LLD)

Bridging Backend Engineering with AI: The Rise of the Intelligent Platform Engineer

10 Steps To Digital Transformation While Simultaneously Cutting Costs

12 Tips to Help You Manage Your WFH Team

How to Write a Full-Stack Application to a Backend Engineer

Why Would Anyone Call a Race Condition Nice?

The In‑Depth, Crystal‑Clear Roadmap to Learning System Design End‑to‑End (HLD & LLD)

Bridging Backend Engineering with AI: The Rise of the Intelligent Platform Engineer

10 Steps To Digital Transformation While Simultaneously Cutting Costs

12 Tips to Help You Manage Your WFH Team

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps