Tech debt rewrite ROI calculator

When tech debt rewrites pay back — and when they don't

Rewrites get pitched as inevitable, but the math rarely supports them. Most rewrites: take 2-3x estimated time, deliver 10-20% velocity gain (not the promised 50%+), break in production, lose institutional knowledge. The good news: well-executed rewrites of the right code DO pay back, often in <12 months.

The math

rewrite cost = eng-months × fully-loaded cost
monthly gain = current spend × velocity gain + current spend × bug time × bug reduction
payback months = rewrite cost ÷ monthly gain

Default scenario: 8 eng-months × $16.5k = $132k cost. Monthly gain = $165k × 35% + $165k × 25% × 40% = $74k/mo. Payback in 1.8 months. 4-year ROI: $3.4M.

Why most rewrites underperform

• Velocity gain assumed too high: 50%+ promised, 10-20% delivered
• Cost underestimated 2x: 8 months becomes 16
• Knowledge loss: rewrite team doesn't have full domain context
• Migration tax: 6+ months running both systems in parallel
• Feature freeze: new features delayed during rewrite, opportunity cost

When rewrites work

• Clear bottleneck: specific service is the gating factor for everything
• Modern foundation gap: codebase >5 years old, dependencies past EOL
• Team retention crisis: top engineers threatening to leave over codebase quality
• Architectural shift: monolith → microservices makes economic sense
• Performance ceiling: current architecture can't scale to next 12-month traffic

The "strangler fig" alternative

Instead of full rewrite, gradually replace parts:

1. Add new functionality in new code; integrate via API
2. Migrate features one at a time as they need updates
3. Eventually old code is just glue or deprecated

Lower risk, slower, retains institutional knowledge, often the right path.