Documents/analysis/Assessment Strategy

Assessment Strategy

Submission Strategy — Approach & Methodology

This is a "meta-document" — how we THINK about the assessment, not the submission itself


1. What Does the Assignment Actually Ask For?

1.1 Deliverables Checklist

# Deliverable Weight Difficulty
4.1 Target Architecture Overview ★★★★★ Hard — this is the core section
4.2 Migration Strategy ★★★★★ Hard — 4+ phases, zero downtime
4.3 Failure Modeling ★★★☆☆ Medium — 5 scenarios + matrix
4.4 Trade-Off Log ★★★★☆ Medium — but shows maturity
4.5 Assumptions ★★☆☆☆ Easy — 8+ items
4.6 AI Usage Declaration ★★★★☆ Easy write, HIGH strategic value

1.2 Hard Constraints of the Submission

Format: PDF
Length: UNDER 6 PAGES
AI Declaration: MANDATORY

6 pages = ~3000-3600 words (depending on font, diagrams)
6 deliverables in 6 pages = ~1 page per section

→ MUST be extremely concise. Tables + diagrams > prose.

2. What Is the Assessor Actually Testing?

2.1 Skills Being Evaluated

This assessment does NOT test:
  ✗ Coding ability
  ✗ Quiz knowledge (what is microservices?)
  ✗ Strict adherence to a specific methodology

This assessment DOES test:
  ✓ JUDGMENT — making the right calls under constraint pressure
  ✓ COMMUNICATION — articulating architecture clearly and concisely
  ✓ TRADE-OFF THINKING — knowing what to sacrifice, what to keep, and why
  ✓ AI FLUENCY — using AI like a Tech Lead, not like a junior
  ✓ PRAGMATISM — solutions that are feasible, not theoretical
  ✓ RISK AWARENESS — identifying problems before they happen

2.2 Assessor Profile (Inferred)

The person who wrote this assessment is likely:
  • Senior/Staff Engineer or Engineering Manager at PhoenixDX
  • Deep understanding of DDD, microservices, .NET ecosystem
  • Cares about AI-first (this is the company's identity)
  • Has seen many submissions → knows the pattern of "AI wrote it all" sameness
  
They will:
  • Skim for 2 minutes first → first impression matters
  • Look at diagrams first → must be clear, self-explanatory
  • Scrutinize trade-offs → this is what separates junior from senior
  • Examine AI declaration → transparency > perfection
  • Cross-check assumptions vs. the rest → are assumptions consistent?

3. Strategy Composition — 6-Page Layout

3.1 Page Budget

Page 1:  Architecture Overview (diagram + service boundaries)
Page 2:  Architecture Details (communication model, data strategy)
Page 3:  Migration Strategy (4 phases, timeline, zero-downtime approach)
Page 4:  Failure Modeling (5 risks table) + Trade-Off Log
Page 5:  Assumptions + AI Declaration
Page 6:  Appendix — glossary, references (if needed, or use for overflow)

Alternative layout (denser):
Page 1:  Executive Summary + Architecture Diagram
Page 2:  Service Boundaries + Communication Model
Page 3:  Migration Phases (timeline visual)
Page 4:  Zero-Downtime Strategy + Failure Modeling
Page 5:  Trade-Offs + Assumptions
Page 6:  AI Declaration + Key Decisions Summary

3.2 Writing Principles

Rule Why
Diagrams first, text second Assessor will look at diagrams first. 1 diagram = 500 words
Tables over paragraphs Scannable, dense, professional
Bold keywords Assessor skims → bold = attention anchor
No fluff "In today's rapidly evolving landscape..." → DELETE
Show judgment "We chose X over Y because Z" > "We use X"
Explicit trade-offs Every decision = benefit + cost stated
Number everything Phase 1, 2, 3, 4. Risk 1-5. Assumption 1-8. Easy to reference

4. Tactics Per Deliverable

4.1 Target Architecture Overview (Page 1-2)

Tactic: Lead with diagram

Content:
  1. ASCII/clean diagram — full system view
     • Frontend → API Gateway → Services → DBs
     • Events bus connecting services
     • ACL → Legacy monolith
     
  2. Service boundaries table
     | Service | Bounded Context | Owner | Phase |
     
  3. Communication model
     • Sync: REST via YARP gateway  
     • Async: Azure Service Bus domain events
     • Pattern: CQRS + Event-driven

Wow factor: Diagram must be clean + complete. If the reader only looks at one thing → it's the diagram.

4.2 Migration Strategy (Page 3)

Tactic: Timeline visual + phase table

Content:
  1. Phased timeline (gantt-style ASCII)
     Phase 0 (M1): Foundation
     Phase 1 (M2-4): Core extraction
     Phase 2 (M5-7): Remaining modules
     Phase 3 (M8-9): Stabilize + optimize
     
  2. Per-phase table
     | Phase | Months | Extract | Key Milestone |
     
  3. Zero-downtime approach (brief)
     • Strangler Fig via YARP routing
     • Canary → gradual traffic shift
     • Instant rollback capability
     • Feature flags per module

Wow factor: Crystal clear WHAT happens WHEN. Assessor sees the timeline and immediately knows you have a feasible plan.

4.3 Failure Modeling (≈ 0.5 page)

Tactic: Dense table, 5 risks

| # | Scenario | Likelihood | Impact | Mitigation |
|---|----------|-----------|--------|------------|
| 1 | Service cascading failure | Medium | Critical | Circuit breaker, bulkhead |
| 2 | Data inconsistency during CDC | High | High | Dual-write validation, reconciliation |
| 3 | Legacy Payment ACL failure | Low | Critical | Retry + DLQ + fallback to direct call |
| 4 | Team key-person dependency | Medium | High | Cross-training, shared ownership |
| 5 | AI-generated code defect in prod | Medium | Medium | Mandatory review gate, contract tests |

Wow factor: Risk #5 = AI-aware risk. Assessor at an AI-first company will appreciate this.

4.4 Trade-Off Log (≈ 0.5 page)

Tactic: 3 columns — chose/over/because

| Decision | Chose (and trade-off) | Over (alternative) |
|----------|---------------------|---------------------|
| Hosting | Container Apps (less control) | AKS (too much ops for 5 eng) |
| DB | Azure SQL everywhere (not optimized) | Polyglot (too much expertise) |
| Payment | Frozen Phase 1 (tech debt: ACL) | Migrate early (too risky) |
| Testing | Contract tests (not E2E heavy) | E2E (slow, flaky, expensive) |
| Frontend | Single SPA (not micro-FE) | Micro-frontends (overkill) |

Wow factor: Explicitly naming WHAT you sacrifice. Shows maturity.

4.5 Assumptions (≈ 0.3 page)

Tactic: Numbered list, bold assumption, brief justification

1. **Legacy codebase has minimal test coverage** → AI generates characterization tests
2. **Team has .NET 8 experience** → no major re-skilling needed
3. **Azure cloud environment available** → no cloud migration overhead
4. **Legacy DB is SQL Server** → CDC tools available (Debezium, built-in CDC)
5. **No regulatory/compliance changes during migration** → scope stable
6. **Product owner available for domain clarification** → no requirement gaps
7. **Legacy monolith APIs are documented or discoverable** → AI can analyze
8. **Budget approved for AI tooling licenses** → Cursor Pro, CodeRabbit, etc.
9. **No mobile app in scope** → web-only modernization
10. **SLA target: 99.9% availability** → standard enterprise, not five-9s

Wow factor: 10 assumptions (exceeds the 8 requirement). Practical, not theoretical.

4.6 AI Usage Declaration (≈ 0.5 page)

Tactic: Transparent + strategic. This is the DIFFERENTIATING section.

Structure:
  1. AI Tools Used
     • Claude (analysis, document generation)
     • Cursor Pro (architecture exploration)
     
  2. AI-Assisted Sections  
     • All sections used AI for initial drafts
     • Architecture diagram: AI-generated, manually refined
     • Risk matrix: AI-suggested, manually prioritized
     
  3. Manual Validation
     • All trade-offs are personal engineering judgment
     • Phase timeline validated against team capacity math
     • Service boundaries based on domain analysis, not AI default
     
  4. AI Governance for Backend Team
     • Review gate: all AI code must pass peer review
     • Contract tests: verify service boundaries (Pact)
     • Prompt versioning: track what prompts generate what code
     • Weekly AI effectiveness retro: measure actual vs expected velocity

Wow factor: Section 4 (governance) = you don't just USE AI, you know how to GOVERN AI usage across the team.
This is Tech Lead thinking, not developer thinking.

5. Mistakes to Avoid

# Mistake Consequence How to Avoid
1 Exceeding 6 pages Auto-fail or bad impression Outline first, strict page budget
2 Overly complex diagrams Assessor skips, misses main point 1 main diagram, max 2 detail diagrams
3 Listing tech stack without justification Looks like copy-paste EVERY choice must have a "because"
4 Over-promising Assessor knows 5 eng + 9 months limits Explicit defer list, realistic scope
5 AI declaration says "AI only assisted" Lacks transparency State clearly which parts are AI-heavy, which are manual judgment
6 Ignoring Payment frozen constraint Misses key constraint ACL + frozen until Phase 3 stated explicitly
7 No rollback plan Zero downtime becomes lip service Every phase must have a rollback mechanism
8 Generic microservices advice Not specific to the assignment Reference 40K users, 5 eng, 9 months specifically

6. Execution Timeline

Step 1: Domain Analysis (done — Business Domain.md)
  └── Understand bounded contexts, coupling points
  
Step 2: Architecture Decision (done — Architect.md, Analysis v2.md)
  └── Service boundaries, communication model, data strategy
  
Step 3: Migration Plan (done — Planning.md)
  └── 4 phases, timeline, team allocation
  
Step 4: Risk + Trade-offs (done — scattered across docs)
  └── Consolidate into Failure Modeling + Trade-Off Log
  
Step 5: Draft Submission (Submission.md)
  └── Compose 6 pages from existing materials
  
Step 6: Review + Compress
  └── Cut to exactly ≤6 pages
  └── Ensure every page earns its space
  └── Check consistency (numbers, phases, names)
  
Step 7: Export PDF
  └── Format check: diagrams render, tables align
  └── Final proofread

7. Competitive Edge — What Makes This Submission Stand Out?

Level 1 (Average):   Migration plan + architecture diagram
Level 2 (Good):      + Clear trade-offs + realistic timeline
Level 3 (Excellent): + AI as strategy (not tool)
                     + Capacity math (44 effective MM)
                     + Constraint interaction analysis
                     + Governance framework for AI in team
                     + Explicit defer list (maturity signal)
                     + Domain-driven decomposition (not module-mapping)

Target: Level 3.
Key differentiator: AI is not just a tool you use — 
  it is a STRATEGY you design for the team.