The Wrong Tech Stack Decision That Cost 6 Months

Context

Plutus pivoted from mobile app to web based on user research showing customers preferred web over installing another app. To minimize rewrite cost, the team reused the Flutter codebase via Flutter Web—optimizing for short-term velocity while ignoring web-specific constraints.

My Role: I joined after the initial decision and led the strategic reset and Next.js migration.

What Went Wrong

Performance

Initial bootstrap: >5 seconds
Data fetch + processing: 5-7 seconds
Total time-to-interactive: 10-12 seconds
Result: High bounce rates and poor engagement

Architecture

Flutter Web rendered via canvas, not HTML
SEO effectively unusable for creator-driven discovery
Accessibility constraints (non-semantic rendering)
Third-party dependencies incompatible with Flutter's renderer

Engineering Cost

~3 months invested in Flutter Web
Optimization attempts delivered <20% improvement
No path to sub-3s TTI without fundamental rewrite

Strategic Reset

I pushed for a complete rewrite guided by one principle:

If we're paying the rewrite cost anyway, the outcome must buy long-term leverage—not just fix the current problem.

Framework Decision: Next.js

Rationale:

Native HTML rendering with SSR/SSG
Predictable performance characteristics
Strong ecosystem and hiring pool
Clear SEO and accessibility path

Platform Hardening

Used the rewrite window to upgrade delivery maturity:

CI/CD Implementation

GitHub Actions with automated testing
Staging auto-deployed from main
Production gated via locked prod branch
Blue-green deployments with instant rollback

Infrastructure Improvements

AWS CloudFront CDN with edge caching
Anti-bot protections and DDoS mitigation
Zero-downtime deployment capability

Results

| Metric | Flutter Web | Next.js | Improvement | | --- | --- | --- | --- | | Time-to-Interactive | 10-12s | <3s | ~75% reduction | | SEO Capability | Blocked | Fully functional | ∞ | | Deployment Time | Manual, 30+ min | Automated, 5 min | 83% reduction | | Production Incidents | Common | Near zero | >90% reduction | | Total Time Cost | 3 months (abandoned) | 3 months (delivered) | ~6 months lost |

Key Lessons

Code reuse is not a strategy — Reusing code increases total cost if the abstraction is wrong
Stack mistakes compound non-linearly — Every feature inherits the performance and delivery debt
Web products require web-native primitives — Canvas-based rendering introduces structural ceilings
Rewrites are leverage moments — Also fix delivery, security, and operability
Platform decisions directly impact runway — This cost ~6 months + opportunity cost

Consulting Takeaway

Seed-stage teams rarely fail from bad features—they fail from early architectural decisions that silently tax every release. My role as consultant is to challenge stack decisions before they ossify and ensure pivots improve speed, reliability, and optionality.

Completed as CTO during pre-seed → seed transition