One tech decision I’d take back involves over-engineering a system early on by embracing microservices architecture far too soon for a startup-scale project.
❌ The Decision
I helped design a backend system split into multiple microservices (auth, notifications, user profiles, analytics, etc.) before we had product-market fit or even five developers on the team.
At the time, it felt right—modular, scalable, future-proof. All the blog posts and conference talks were singing the praises of microservices. But in practice:
We spent more time managing inter-service communication, deployments, and failure points than building actual features.
Debugging was a nightmare—issues crossed service boundaries constantly.
Dev velocity tanked because even small changes required coordination across multiple services and repos.
✅ The Lesson
Premature optimization is not just about speed—it's also about complexity.
Start simple. Monoliths aren’t bad when you're still validating your product or have a small team. You can always refactor later, and it’s far easier to split a well-structured monolith than it is to untangle a web of services that were rushed into place.