Two years on autopilot, a working FinTech product, zero maintenance. Until the day everything collapsed.
QuickFund was profitable from month one. Micro-loans, automatic repayments, homemade scoring. It worked. So I stopped touching it. For two years. That's the worst technical decision I ever made.
€50K in losses because I refused to maintain code that was running. Not a hack, not the market. Just technical laziness.
Incident timeline
QuickFund launch
Working MVP. Basic scoring, minimal interface, semi-automated process. Profitable from month 1.
Autopilot mode
The product runs, I focus on other subsidiaries. Zero updates, zero refactoring, zero serious monitoring.
First signals
A few bugs reported by clients. Response times increasing. I patch urgently, never fix the root cause.
The incident
Critical data loss. Repayments not recorded. Clients billed twice. Scoring returning inconsistent results.
Radical decision
I stop everything. Full stop. No new loans. Complete audit of the code, database, and processes.
Restructuring
Complete rewrite. New stack, new scoring, new processes. Today's QuickFund has nothing in common with the initial MVP.
Loss breakdown
Lost repayments
Payments received but not recorded in the system
Double billing
Clients charged twice, manual refunds
Unrecoverable loans
Faulty scoring approving risky profiles
Restructuring cost
Time, tools, external audit
Total
What caused the problem
Technical debt is code you know is fragile but don't fix. Tests you don't write. Monitoring you don't set up because "it works." On QuickFund I had accumulated:
Zero automated tests — every deployment was a gamble
Database without integrity constraints — duplicates everywhere
No logging system — when the bug hits, you don't know when or why
API without versioning — one change broke everything silently
No automated backup — the database was the single point of failure
What I did
The hardest decision is stopping a product that's making money. Telling clients I'm suspending new loans. But it was that or keep patching on rotten code.
Complete audit — 3 weeks to map every component
Scoring rewrite — algorithm reviewed, tested, manually validated
Database migration — clean PostgreSQL with constraints and indexing
Test implementation — 85%+ coverage before any redeployment
Real-time monitoring — alerts on every financial anomaly
Deployment process — CI/CD, staging, mandatory human validation
The lessons
Five rules I now apply in every company in the group.
A running product isn't a maintained product
Working today guarantees nothing for tomorrow. Code ages, dependencies too.
No code in prod without tests
If you can't test it, you can't deploy it.
Monitoring isn't optional
You need to know in real-time what's happening. Especially when you're handling money.
20% of dev time = maintenance
If you don't plan maintenance, the incident plans it for you.
Cutting beats patching
When the debt is too deep, you need the courage to stop. Not keep duct-taping.
The takeaway
€50K because I thought "it runs" meant "it's solid." It was just luck. Today QuickFund is more robust than ever, but I could have avoided all of this with 2 hours a week. Technical debt is a loan on your future self, and the interest is brutal.
GL