Documents/analysis/Cost Analysis

Cost Analysis

Cost Analysis — Modernization Project

Objective: Full cost analysis of the .NET 8 microservices modernization project over 9 months
Principles: Transparent, realistic, worst-case aware — no overselling, no hidden costs
Source: Planning.md, AI Strategy.md, Architect.md, Deployment.md, HA.md, Constraints Analysis.md, Trade-Off Log


1. Executive Summary

┌──────────────────────────────────────────────────────────────────┐
│  TOTAL PROJECT COST ESTIMATE: $815K – $1.13M (9 months)         │
│                                                                  │
│  ┌────────────────────────────────────────────────────┐          │
│  │  Team (salaries)         $500K–750K    (62–66%)    │          │
│  │  Azure Infrastructure    $150K–250K    (18–22%)    │          │
│  │  AI Tooling              $7K–10K       (~1%)       │          │
│  │  Testing/QA Tools        $10K–20K      (1–2%)      │          │
│  │  Training/Onboarding     $5K–10K       (<1%)       │          │
│  │  Contingency (10%)       $50K–100K     (6–9%)      │          │
│  └────────────────────────────────────────────────────┘          │
│                                                                  │
│  Cost per effective man-month: ~$18.5K–25.7K/MM                  │
│  AI ROI: $7K–10K investment → +17 MM output (2x multiplier)     │
│  Break-even: Month 2 (AI cost recovered by Month 2 productivity)│
└──────────────────────────────────────────────────────────────────┘

Why cost analysis matters for the assessment?

  • Assessors want to see that the Tech Lead understands the business impact of technical decisions
  • Every architecture choice = cost choice. Container Apps vs AKS = $50K+ difference
  • AI tooling cost must be justified by ROI, not because "it's cool"

2. Team Cost (Largest Component: 62–66%)

2.1 Salary Estimation

Role Engineer Estimated Monthly (VN Hub) 9 Months
Tech Lead / Architect D1 $4,000–6,000 $36K–54K
Senior Backend D2 $3,000–5,000 $27K–45K
Mid Backend D3 $2,500–4,000 $22.5K–36K
Fullstack D4 $2,500–4,000 $22.5K–36K
Frontend + DevOps D5 $2,500–4,000 $22.5K–36K
TOTAL 5 $14.5K–23K/mo $130.5K–207K

Note: Salary range depends on the Vietnam market (HCM/Hanoi) vs global remote rates. The table above uses Vietnam Hub rates. For mixed global teams, multiply by ×2–3.

2.2 Loaded Cost (Salary + Benefits + Overhead)

Base salary:                    $130K–207K
Benefits (insurance, bonus):    × 1.3–1.5 multiplier
Office/equipment:               $2K/person × 5 = $10K
──────────────────────────────────────────────────
Loaded team cost:               $180K–320K (9 months, VN rates)
                            or  $500K–750K (if global benchmark)

2.3 Capacity vs Cost Efficiency

Metric Traditional (1.0x) AI-Heavy (2.0x)
Raw capacity 45 MM 45 MM
Overhead deduction -18 MM (40%) -18 MM (40%)
AI setup investment 0 -5 MM (Phase 0)
Net productive MM 27 MM Phase-by-phase: 45.5 ≈ ~44 MM (variable AI multiplier per phase)
Cost per effective MM $18.5K/MM $11.4K/MM
Output equivalent 5 engineers ~7.5 engineers

Insight: AI investment does not significantly increase cost but boosts output by 63%.


3. Azure Infrastructure Cost

3.1 Service Inventory

┌─────────────────────────────────────────────────────────────────┐
│  AZURE RESOURCE MAP (Production)                                 │
│                                                                  │
│  ┌─ Azure Front Door ──────────────────────────────┐            │
│  │  Global load balancing + CDN + WAF              │            │
│  │  Est: $35–50/month (Standard tier)              │            │
│  └────────────────────┬────────────────────────────┘            │
│                       │                                          │
│  ┌────────────────────▼────────────────────────────┐            │
│  │  YARP API Gateway (Container App)               │            │
│  │  Est: $30–60/month (0.5 vCPU, 1 GiB)           │            │
│  └────────────────────┬────────────────────────────┘            │
│                       │                                          │
│  ┌────────┬───────────┼───────────┬────────────────┐            │
│  │        │           │           │                │            │
│  ▼        ▼           ▼           ▼                ▼            │
│ Travel  Event    Workforce     Comms         Reporting          │
│ $60/mo  $45/mo    $45/mo      $30/mo          $30/mo           │
│ (1vCPU) (0.75)   (0.75)      (0.5)           (0.5)            │
│                                                                  │
│  ┌─────────────────────────────────────────────────┐            │
│  │  Azure SQL (per-service databases)              │            │
│  │  5 DBs × Standard S2 (50 DTU) = $75/db/mo      │            │
│  │  Total: $375/month                              │            │
│  └─────────────────────────────────────────────────┘            │
│                                                                  │
│  ┌─────────────────────────────────────────────────┐            │
│  │  Azure Service Bus (Standard tier)              │            │
│  │  Base: $10/month + $0.05 per million ops        │            │
│  │  Est: $15–25/month (40K users)                  │            │
│  └─────────────────────────────────────────────────┘            │
│                                                                  │
│  Shared: ACR ($5/mo), Key Vault ($5/mo),                        │
│          App Insights (~$45/mo for 20GB logs)                   │
└─────────────────────────────────────────────────────────────────┘

3.2 Monthly Cost Breakdown (Per Environment)

Service Dev Staging Production Notes
Container Apps (6 services) $50 $100 $250 Dev: min replicas. Prod: auto-scale 2–10
Azure SQL (5 databases) $125 $250 $375 Dev: Basic (5 DTU). Staging: S1. Prod: S2 (50 DTU)
Service Bus $10 $10 $25 Standard tier. Prod: higher message throughput
Azure Front Door + CDN $50 Prod only. Standard tier + WAF
Container Registry $5 Shared across envs. Basic tier
Key Vault $2 $2 $2 Minimal cost per env
App Insights / Monitor $10 $15 $45 Based on log/trace volume
Azure DNS $5 Hosted zone + queries
TOTAL/month $202 $377 $752

3.3 Infrastructure Cost Over 9 Months

Phase 0 (M1):     Dev only                    = $202
Phase 1 (M2-4):   Dev + Staging + Prod ramp   = $202 + $377 + $400 = $979/mo × 3 = $2,937
Phase 2 (M5-7):   All envs, full prod         = $202 + $377 + $752 = $1,331/mo × 3 = $3,993
Phase 3 (M8-9):   All envs, stable            = $1,331/mo × 2 = $2,662
────────────────────────────────────────────────────────────────────
TOTAL INFRASTRUCTURE (9 months):               = ~$9,794

Annualized run rate (post-project):            = ~$16K/year

Note: Estimates based on Azure public pricing (Southeast Asia region, March 2026). Actual cost depends on Reserved Instance discounts, Enterprise Agreement, and real traffic patterns.

3.4 Infrastructure Cost Comparison: Choices We Made

Decision Our Choice Alternative Cost Difference Why
Container orchestration Container Apps AKS (Kubernetes) Saves $200–400/mo AKS: min 3-node cluster × $100/node + ops engineer time. Container Apps: pay-per-use, no cluster
Database Azure SQL (all) Polyglot (Cosmos + Redis + SQL) Saves $300–500/mo Cosmos DB: $25/100 RU/s × 5 services = expensive. Redis: $60/mo per instance. One DB tech = simpler
Region Single (SEA) Multi-region active Saves $800–1,200/mo Active-active = 2× compute, 2× DB replication, 2× Service Bus
Message broker Service Bus Standard Kafka (Event Hubs) Saves $100–200/mo Kafka: min $350/mo (Event Hubs Standard). Service Bus: $10/mo base
Gateway YARP (Container App) Azure API Management Saves $150–300/mo APIM Standard: $150+/mo. YARP = code running in Container App ($30/mo)

Total savings from architecture choices: $1,550–2,600/month compared to the "enterprise default" stack.


4. AI Tooling Cost

4.1 Monthly Tool Cost

Tool Per Engineer Team (5) Monthly Cost 9-Month Cost
Cursor Pro $20 5 $100 $900
Claude API (agentic usage) $80–150 5 $400–750 $3,600–6,750
Claude.ai Pro $20 5 $100 $900
ChatGPT Plus $20 5 $100 $900
CodeRabbit $15/seat 5 $75 $675
Ollama (local models) $0 5 $0 $0
TOTAL $775–1,125/mo $6,975–10,125

4.2 AI ROI Analysis

┌────────────────────────────────────────────────────────────────┐
│  AI INVESTMENT vs RETURN                                        │
│                                                                  │
│  Investment:                                                     │
│    AI tools (9 months):           $7K–10K                       │
│    Phase 0 setup (5 MM × $3K):    $15K                          │
│    ────────────────────────────                                 │
│    Total AI investment:           $22K–25K                       │
│                                                                  │
│  Return:                                                         │
│    Additional output:             +17 MM (from 27 → 44 MM)     │
│    Value of 17 MM (@ $11K/MM):    $187K                         │
│    ────────────────────────────                                 │
│    NET VALUE:                     $162K–165K                     │
│    ROI:                           ~700%                          │
│                                                                  │
│  Break-even:                                                     │
│    Month 2 (first full month of AI-assisted development)        │
│    By Month 2, AI tools have saved >$25K in equivalent output   │
│                                                                  │
│  Risk if AI tools NOT approved:                                  │
│    Capacity drops from 44 MM → 27 MM                            │
│    Must cut 2–3 services from scope                             │
│    Timeline extends 9 months → 14 months                        │
│    Cost increase: +$125K–225K (5 extra months of team salary)   │
└────────────────────────────────────────────────────────────────┘

4.3 AI Cost per Task Category

Task Type AI Multiplier AI Tool Primarily Used Cost Attribution
Boilerplate/scaffolding 5.0x Cursor Pro (autocomplete) $100/mo (lowest)
Business logic migration 1.5x Claude API (agentic) $750/mo (highest — complex prompts)
Test generation 3.0x Claude API + Cursor $400/mo
Documentation 4.0x Claude.ai Pro $100/mo
Code review 2.5x CodeRabbit + Claude $150/mo
DevOps/IaC 2.0x Cursor + Copilot $120/mo

Insight: Claude API accounts for 67% of AI cost but drives 50% of the AI multiplier through business logic migration — the largest investment but also the most valuable.


5. Testing & QA Tool Cost

Tool Type Cost Model 9-Month Estimate
Pact (contract testing) Open-source Free $0
Pactflow (Pact broker) SaaS $30–100/mo (team plan) $270–900
Playwright (E2E testing) Open-source Free $0
SonarCloud (code quality) SaaS Free for OSS / $15/mo private $0–135
CodeQL (SAST) GitHub included Free with GitHub $0
GitHub Actions (CI/CD) GitHub Free tier + $4/min macOS $500–2,000
TOTAL $770–3,035

6. Training & Onboarding Cost

Item Who Method Estimated Cost
.NET 8 migration patterns D2, D3, D4 Pluralsight / internal workshops $500–1,000
AI tool adoption (Phase 0) All 5 Internal training + documentation $0 (time cost only)
Azure Container Apps D5 Microsoft Learn + hands-on $0 (free resources)
DDD / CQRS patterns D3, D4 Internal workshops by D1 $0 (time cost only)
Contract testing (Pact) All Internal workshop + Pact docs $0
TOTAL $500–1,000

Most training cost = time cost (counted in the 40% overhead deduction), not dollar cost.


7. Total Cost Summary

7.1 Consolidated Budget (9 Months)

Category Low Estimate High Estimate % of Total
Team salaries (loaded) $180,000 $320,000 62–66%
Azure infrastructure $9,800 $15,000 3–4%
AI tooling $7,000 $10,200 2–3%
Testing/QA tools $770 $3,000 <1%
Training $500 $1,000 <1%
GitHub (Actions, repos) $500 $2,000 <1%
Contingency (15%) $29,800 $52,700 15%
TOTAL $228,370 $403,900 100%

Range: $228K–$404K at VN Hub rates. Using global market rates ($8K–15K/engineer/month), total project cost rises to $815K–$1.13M.

7.2 Monthly Burn Rate

Monthly Burn Rate Over Project Lifecycle:

Month:  M1      M2      M3      M4      M5      M6      M7      M8      M9
        ┌───┐   ┌───┐   ┌───┐   ┌───┐   ┌───┐   ┌───┐   ┌───┐   ┌───┐   ┌───┐
Team:   │$20K│   │$20K│   │$20K│   │$20K│   │$20K│   │$20K│   │$20K│   │$20K│   │$20K│
        ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤
Azure:  │$0.2│   │$1K │   │$1K │   │$1K │   │$1.3│   │$1.3│   │$1.3│   │$1.3│   │$1.3│
        ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤   ├───┤
AI:     │$1K │   │$1K │   │$1K │   │$1K │   │$1K │   │$1K │   │$1K │   │$1K │   │$1K │
        └───┘   └───┘   └───┘   └───┘   └───┘   └───┘   └───┘   └───┘   └───┘
Total:  $21.2K  $22K    $22K    $22K    $22.3K  $22.3K  $22.3K  $22.3K  $22.3K

Cumulative: $21K → $43K → $65K → $87K → $110K → $132K → $154K → $177K → $199K

7.3 Cost per Delivered Value

Metric Value
Total cost (VN rates) ~$228K–404K
Effective man-months 44 MM
Cost per effective MM $5.2K–9.2K/MM
Services delivered 5 (Travel, Event, Workforce, Comms, Reporting) + ACL + Gateway
Cost per service $32.6K–57.7K/service
Cost per active user (40K) $5.7–10.1/user (one-time migration)

8. Cost Risks & Mitigation

Risk Impact Likelihood Mitigation
Azure cost overrun (traffic spike, misconfigured auto-scale) +$2K–5K/mo Medium Budget alerts at 80%/100%. Auto-scale ceiling = 10 replicas. Review weekly
AI tool price increase (Claude API pricing change) +$200–500/mo Low Fallback to local models (Ollama). Cap API spend per engineer
Scope creep → extended timeline +$20K/month per extra month Medium Strict phase gates. Defer to Phase 4 if scope grows. Weekly burn rate review
Multi-region required earlier +$800–1,200/mo Low Only if user complaints > 5% about latency. CDN mitigates most cases
AI multiplier underperforms (1.5x instead of 2.0x) -6 MM capacity → cut 1 service Medium Phase 0 pilot measures actual multiplier. Adjust scope at Phase 1 gate
Key engineer departure (D2 leaves) +$15K–30K recruitment + 2 month ramp Medium Cross-training from M2. Knowledge in git, not in heads

Cost Containment Rules

RULE 1: No Azure service above Standard tier unless proven bottleneck
RULE 2: AI tool spend capped at $1,200/month. Over = review with team lead
RULE 3: No additional engineer mid-project (Brooks's Law)
RULE 4: Multi-region deferred until user complaint data justifies
RULE 5: Weekly Azure Cost Dashboard review (15 min in standup)
RULE 6: Reserved Instances purchased ONLY after 3 months stable usage
RULE 7: Dev environment auto-shutdown (nights/weekends) — saves 60% dev cost

9. Post-Project Run Cost (Steady State)

Category Monthly Annual
Azure production $752 $9,024
Azure staging $377 $4,524
Azure dev $202 $2,424
AI tooling (reduced team) $400 $4,800
Monitoring / observability $45 $540
GitHub / CI/CD $100 $1,200
TOTAL RUN COST $1,876/mo $22,512/year
Savings vs running legacy monolith:
  
  Legacy: Single VM (large) + single SQL DB + manual deploy
  Typical cost: $500–1,000/month
  
  New: Microservices (auto-scale) + 5 DBs + managed services  
  Cost: $1,876/month
  
  Cost increase: +$876–1,376/month for production
  But: auto-scale down during low traffic → actual: +$400–800/month
  
  TRADE-OFF: Higher infra cost → but:
    ✓ Zero downtime deployment (no maintenance windows)
    ✓ Independent team scaling (each service → different team)
    ✓ Per-service auto-heal (one crash ≠ system down)
    ✓ Feature velocity 3–5x faster (independent deploys)
    
  At 40K users, $22K/year = $0.55/user/year for modern infra
  → ACCEPTABLE for enterprise travel/event platform

10. Decision Matrix: What Drives Cost

COST SENSITIVITY ANALYSIS

                    High Impact on Cost
                           ▲
                           │
        ┌──────────────────┼──────────────────┐
        │                  │                  │
        │  Timeline ext.   │  Multi-region    │
        │  (+$20K/month)   │  (+$10K/year)    │
        │                  │                  │
  Easy ─┼──────────────────┼──────────────────┼─ Hard
  to    │                  │                  │  to
  control│  AI tool choice  │  Team attrition  │  control
        │  (switch models) │  (market risk)   │
        │                  │                  │
        └──────────────────┼──────────────────┘
                           │
                    Low Impact on Cost
                           ▼

KEY INSIGHT: Biggest cost risk = timeline extension, not infrastructure.
One extra month = $22K. 
All Azure infra for 9 months = $10K.

→ OPTIMIZE FOR SPEED, NOT FOR INFRA COST.
→ AI investment ($10K) that saves 1 month ($22K) = net positive.

11. Cost Comparison: Build vs Buy vs Hybrid

Approach 9-Month Cost Capacity Risk
Full outsource $400K–800K Depends on vendor High: knowledge transfer, quality control
Hire more engineers (10 instead of 5) $450K–650K 54 MM Medium: Brooks's Law, recruiting time
5 engineers, no AI $180K–320K 27 MM High: scope cut 40%, timeline risk
5 engineers + AI (our approach) $228K–404K 44 MM Low-Medium: proven tools, measurable

Our approach delivers the best cost-per-MM at lowest execution risk.


12. Appendix: Azure Pricing References

Prices based on Southeast Asia (Singapore) region, Pay-As-You-Go, as of March 2026.

Service Tier Unit Price Source
Container Apps Consumption $0.000024/vCPU-s + $0.000003/GiB-s azure.microsoft.com/pricing/details/container-apps
Azure SQL Standard S2 (50 DTU) ~$75/month azure.microsoft.com/pricing/details/azure-sql-database
Service Bus Standard $0.05/mil operations + $10/mo base azure.microsoft.com/pricing/details/service-bus
Azure Front Door Standard $35/mo + per-request azure.microsoft.com/pricing/details/frontdoor
Container Registry Basic $5/mo azure.microsoft.com/pricing/details/container-registry
Key Vault Standard $0.03/10K operations azure.microsoft.com/pricing/details/key-vault
App Insights Pay-as-you-go $2.99/GB (first 5 GB free) azure.microsoft.com/pricing/details/monitor

Disclaimer: Prices fluctuate. Enterprise Agreement (EA) or Reserved Instances can reduce costs by 30–50%. Estimates here use Pay-As-You-Go for worst-case planning.