π§ Tech Lead Decision Making β Quy TrΓ¬nh Ra QuyαΊΏt Δα»nh Hiα»u QuαΊ£ (70% Rule, Trade-offs & Ownership)
π― Mα»₯c TiΓͺu BΓ i ViαΊΏt
Mα»t trong nhα»―ng kα»Ή nΔng quan trα»ng nhαΊ₯t của Tech Lead lΓ ra quyαΊΏt Δα»nh. BΓ i nΓ y Δi sΓ’u vΓ o quy trΓ¬nh decision making, cΓ‘ch phΓ’n tΓch trade-offs, trΓ‘nh analysis paralysis, vΓ mindset ownership khi quyαΊΏt Δα»nh sai.
Good decision now > Perfect decision too late.
Series Navigation
BΓ i 1 β How to Become a Tech Lead (Career Path & Mindset)
BΓ i 2 β Engineering Manager (Technical Skill & Leadership)
BΓ i 3 β Tech Lead Time Management (40% Coding, 25% Meetings...)
BΓ i 4 β (bΓ i nΓ y) Tech Lead Decision Making
π 1. Tech Lead Decision Making Flow β Tα»ng Quan
Quy TrΓ¬nh 5 BΖ°α»c
PROBLEM / DECISION NEEDED
β
βΌ
βββββββββββββββββββββββββ
β 1. Collect Informationβ
β (Δủ, khΓ΄ng cαΊ§n hoΓ n hαΊ£o)β
βββββββββββββ¬ββββββββββββ
β
βΌ
βββββββββββββββββββββββββ
β 2. Analyze Options β
β + Trade-offs β
βββββββββββββ¬ββββββββββββ
β
βΌ
βββββββββββββββββββββββββ
β 3. Make Decision β
β (KhΓ΄ng trΓ¬ hoΓ£n) β
βββββββββββββ¬ββββββββββββ
β
βΌ
βββββββββββββββββββββββββ
β 4. Execute Plan β
β Implement solution β
βββββββββββββ¬ββββββββββββ
β
βΌ
βββββββββββββββββββββββββ
β 5. Evaluate Result β
βββββββββββββ¬ββββββββββββ
β
ββββββββββββββ΄βββββββββββββ
βΌ βΌ
ββββββββββββββββββ ββββββββββββββββββ
β Decision Right β β Decision Wrong β
β β Continue β β β Learn & Adjustβ
β β Document β β β No blame β
ββββββββββββββββββ ββββββββββββββββββ
Key insight: Quy trΓ¬nh nΓ y lΓ loop, khΓ΄ng phαΊ£i one-time. Decision sai β Learn β Adjust β Better decision next time.
π 2. Step 1 β Collect Information
Nguα»n ThΓ΄ng Tin
Information Sources
β
βββββββββββββββββΌββββββββββββββββ
βΌ βΌ βΌ
βββββββββββ βββββββββββ βββββββββββ
βEngineersβ β Product β β Data β
β(technical)β β(business)β β(metrics)β
ββββββ¬βββββ ββββββ¬βββββ ββββββ¬βββββ
β β β
βΌ βΌ βΌ
Technical Requirements Performance
constraints & deadlines & analytics
ThΓ΄ng Tin CαΊ§n Thu ThαΊp
ββββββββββββββββββββββββββββββββββββββββββββββββ
β Information Checklist β
ββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β Technical: β
β ββ System performance metrics β
β ββ Current architecture limitations β
β ββ Technical debt level β
β ββ Team's technical capacity β
β β
β Business: β
β ββ Product requirements β
β ββ Deadline & timeline β
β ββ Business impact / priority β
β ββ Budget constraints β
β β
β Team: β
β ββ Team capacity & availability β
β ββ Skill level β
β ββ Ongoing commitments β
β ββ Team's opinion & concerns β
β β
ββββββββββββββββββββββββββββββββββββββββββββββββ
70% Information Rule
βββββββββββββββββββββββββββββββββββββββββββββββ
β β
β 70% Information Rule β
β β
β NαΊΏu bαΊ‘n cΓ³ ~70% thΓ΄ng tin cαΊ§n thiαΊΏt β
β β
β β MAKE THE DECISION β
β β
β Chα» 100% thΓ΄ng tin = chα» mΓ£i mΓ£i β
β β
βββββββββββββββββββββββββββββββββββββββββββββββ
0% 70% 100%
ββββββββββββββββββββββββββββββββββββββ
β
βΌ
Decision zone β
TrΖ°α»c 70% β ThiαΊΏu info, risky
Sau 70% β Diminishing returns, wasting time
βοΈ 3. Step 2 β Analyze Options & Trade-offs
Không Tìm Solution Hoà n Hảo
Tech Lead khΓ΄ng tΓ¬m solution hoΓ n hαΊ£o β mΓ so sΓ‘nh trade-offs giα»―a cΓ‘c options.
Option Analysis Template:
βββββββββββββββββββββββββββββββββββββββββββββββ
β Option A β
βββββββββββββββββββββββββββββββββββββββββββββββ€
β β
Pros: β
β ββ Fast to implement (2 weeks) β
β ββ Low cost β
β ββ Team familiar β
β β
β β Cons: β
β ββ Creates tech debt β
β ββ Not scalable β
β ββ Need refactor in 6 months β
β β
β Risk: Medium β
β Timeline: 2 weeks β
βββββββββββββββββββββββββββββββββββββββββββββββ
βββββββββββββββββββββββββββββββββββββββββββββββ
β Option B β
βββββββββββββββββββββββββββββββββββββββββββββββ€
β β
Pros: β
β ββ Clean architecture β
β ββ Scalable (10x traffic) β
β ββ Long-term maintainable β
β β
β β Cons: β
β ββ Slower (6 weeks) β
β ββ Higher cost β
β ββ Team needs learning β
β β
β Risk: Low β
β Timeline: 6 weeks β
βββββββββββββββββββββββββββββββββββββββββββββββ
βββββββββββββββββββββββββββββββββββββββββββββββ
β Option C β
βββββββββββββββββββββββββββββββββββββββββββββββ€
β β
Pros: β
β ββ Medium time (3 weeks) β
β ββ Acceptable tech debt β
β ββ Balanced approach β
β β
β β Cons: β
β ββ Not perfect β
β ββ May need iteration β
β β
β Risk: Medium-Low β
β Timeline: 3 weeks β
βββββββββββββββββββββββββββββββββββββββββββββββ
Trade-off Matrix
Trade-off Dimensions:
Option A Option B Option C
βββββββββ βββββββββ βββββββββ
Time βββββ ββ ββββ
Cost βββββ ββ ββββ
Scalability ββ βββββ βββ
Maintainability ββ βββββ ββββ
Risk βββ βββββ ββββ
Common Trade-offs
ββββββββββββββββββββββββββββββββββββββββββββββββ
β Common Trade-offs β
ββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β Speed vs Quality β
β ββ Ship fast β more bugs β
β ββ Ship clean β miss deadline β
β β
β Build vs Buy β
β ββ Build β control, time-consuming β
β ββ Buy β fast, dependency β
β β
β Consistency vs Availability (CAP) β
β ββ Strong consistency β slower β
β ββ Eventual consistency β faster, complex β
β β
β Technical Debt vs Delivery β
β ββ No debt β slow delivery β
β ββ Accept debt β need to pay later β
β β
β Generalization vs Specialization β
β ββ Generic β flexible, complex β
β ββ Specific β simple, less reusable β
β β
ββββββββββββββββββββββββββββββββββββββββββββββββ
β±οΈ 4. Step 3 β Make Decision (ΔΓΊng LΓΊc)
Timing Quan Trα»ng
Decision Timing:
Perfect decision
but too late
β
βΌ
βββββββββββββββββββ
β Business impact β
β Deadline missed β
β Team blocked β
β Opportunity lostβ
βββββββββββββββββββ
βββββββββββββββββββββββββββββββββββββββ
Good decision
on time
β
βΌ
βββββββββββββββββββ
β Team unblocked β
β Progress made β
β Can iterate β
β Learn & adjust β
βββββββββββββββββββ
Key Insight
βββββββββββββββββββββββββββββββββββββββββββββββ
β β
β Good decision NOW β
β > β
β Perfect decision TOO LATE β
β β
βββββββββββββββββββββββββββββββββββββββββββββββ
Khi NΓ o NΓͺn QuyαΊΏt Δα»nh?
Decision triggers:
ββ CΓ³ ~70% thΓ΄ng tin cαΊ§n thiαΊΏt
β
ββ Team Δang bα» blocked
β
ββ Deadline Δang ΔαΊΏn gαΊ§n
β
ββ ThΓͺm thΓ΄ng tin sαΊ½ khΓ΄ng thay Δα»i decision
β
ββ Cost of delay > cost of wrong decision
π 5. Step 4 β Execute Plan
Tα»« Decision ΔαΊΏn Action
Decision Made
β
βΌ
Break into tasks
β
βΌ
Delegate to team
β
βΌ
Set milestones
β
βΌ
Track progress
β
βΌ
Unblock issues
VΓ Dα»₯: Refactor Payment Module
Decision: Refactor payment module (Option C)
Timeline: 3 weeks
β
βΌ
Task Breakdown:
Week 1:
ββ Dev A β API interface refactor
ββ Dev B β Database schema update
ββ Tech Lead β Architecture review
Week 2:
ββ Dev A β Payment provider integration
ββ Dev B β Migration script
ββ Dev C β Unit tests
Week 3:
ββ Dev A, B, C β Integration testing
ββ Tech Lead β Code review
ββ Team β Bug fixes & deployment
Communicate Decision
Sau khi quyαΊΏt Δα»nh, Tech Lead phαΊ£i communicate:
WHAT β QuyαΊΏt Δα»nh lΓ gΓ¬?
WHY β TαΊ‘i sao chα»n option nΓ y?
HOW β Implement nhΖ° thαΊΏ nΓ o?
WHEN β Timeline & milestones
WHO β Ai lΓ m gΓ¬?
RISKS β Risks vΓ mitigation plan
π 6. Step 5 β Evaluate Result
Measure Impact
Implementation Done
β
βΌ
Measure Impact
β
βΌ
βββββββββββββββββββββββββββββββββββββ
β Metrics β
βββββββββββββββββββββββββββββββββββββ€
β β
β Performance β
β ββ Response time improved? β
β ββ Throughput increased? β
β β
β Stability β
β ββ Error rate reduced? β
β ββ Uptime improved? β
β β
β Delivery β
β ββ On time? β
β ββ Within budget? β
β ββ Scope met? β
β β
β Team β
β ββ Team learned? β
β ββ Process improved? β
β β
βββββββββββββββββββββββββββββββββββββ
Outcome Handling
Evaluation
β
ββββββββββββ΄βββββββββββ
βΌ βΌ
Decision CORRECT Decision WRONG
β β
βΌ βΌ
Continue Learn & Adjust
Document success No blame culture
Share learnings Post-mortem
Apply to next decision
Khi Decision Sai
Decision sai β KHΓNG PHαΊ’I thαΊ₯t bαΊ‘i
Decision sai + Blame culture:
ββ Team sợ quyαΊΏt Δα»nh
ββ Analysis paralysis
ββ KhΓ΄ng ai dΓ‘m take ownership
Decision sai + Learning culture:
ββ Post-mortem: tαΊ‘i sao sai?
ββ What would we do differently?
ββ Apply learning to next decision
ββ Team grows
π§ 7. Tech Lead Decision Model β Framework
Model Tα»ng Hợp
Tech Lead Decision Model
βββββββββββββββββββ
β Information β (70% rule)
ββββββββββ¬βββββββββ
β
βΌ
βββββββββββββββββββ
β Options β (at least 2-3)
ββββββββββ¬βββββββββ
β
βΌ
βββββββββββββββββββ
β Trade-offs β (pros/cons matrix)
ββββββββββ¬βββββββββ
β
βΌ
βββββββββββββββββββ
β Decision β (timely, not perfect)
ββββββββββ¬βββββββββ
β
βΌ
βββββββββββββββββββ
β Execution β (delegate & track)
ββββββββββ¬βββββββββ
β
βΌ
βββββββββββββββββββ
β Learning β (evaluate & improve)
βββββββββββββββββββ
β
β feedback loop
ββββββββββββββββΊ back to Information
Decision Types & Reversibility
ββββββββββββββββββββββββββββββββββββββββββββββββ
β Decision Reversibility β
ββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β Type 1: Irreversible (one-way door) β
β ββ High impact, hard to undo β
β ββ CαΊ§n nhiα»u analysis hΖ‘n β
β ββ VΓ dα»₯: Architecture thay Δα»i lα»n, β
β Tech stack migration β
β β
β Type 2: Reversible (two-way door) β
β ββ Low impact, easy to undo β
β ββ QuyαΊΏt Δα»nh nhanh β
β ββ VΓ dα»₯: Feature flag, A/B test, β
β Library choice cho 1 module β
β β
ββββββββββββββββββββββββββββββββββββββββββββββββ
Type 1 (Irreversible): Type 2 (Reversible):
Slow down Speed up
More analysis Bias for action
Get more opinions Decide & iterate
Document heavily Can rollback
Jeff Bezos rule: Treat most decisions as Type 2. Only slow down for truly irreversible ones.
π¨ 8. Anti-Patterns β Sai LαΊ§m ThΖ°α»ng GαΊ·p
Analysis Paralysis
β Analysis Paralysis:
Collect info
β
βΌ
Need more info
β
βΌ
Even more info
β
βΌ
Still not enough
β
βΌ
NO DECISION MADE
β
βΌ
βββββββββββββββββββ
β Team blocked β
β Deadline missed β
β Product delayed β
βββββββββββββββββββ
HiPPO (Highest Paid Personβs Opinion)
β HiPPO Decision:
Team discussion
β
βΌ
Data & analysis
β
βΌ
Trade-off comparison
β
βΌ
Boss says "Do X"
β
βΌ
Do X (ignore all analysis)
β Team demotivated
β Bad decisions
β No learning
Decision by Committee
β Decision by Committee:
Everyone must agree
β
βΌ
Endless meetings
β
βΌ
Compromise solution
(worst of all options)
β
βΌ
No one owns it
No one is accountable
Avoiding Decision
β Avoiding Decision:
Problem exists
β
βΌ
"Let's wait and see"
β
βΌ
Problem grows
β
βΌ
"Let's wait more"
β
βΌ
Crisis
β
βΌ
Forced decision (worst timing)
β 9. Best Practices
Decision Documentation
Decision Record Template:
ββββββββββββββββββββββββββββββββββββββββββββββββ
β Decision Record β
ββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β Title: [Decision name] β
β Date: [When decided] β
β Decision Maker: [Who] β
β Status: [Proposed/Accepted/Deprecated] β
β β
β Context: β
β - Problem statement β
β - Constraints β
β β
β Options Considered: β
β - Option A: [pros/cons] β
β - Option B: [pros/cons] β
β - Option C: [pros/cons] β
β β
β Decision: β
β - Chosen option: [which] β
β - Rationale: [why] β
β β
β Consequences: β
β - Positive β
β - Negative (accepted trade-offs) β
β β
ββββββββββββββββββββββββββββββββββββββββββββββββ
Get Input, Make Decision
β
Best Practice:
Gather team input
β
βΌ
Listen to concerns
β
βΌ
Consider all perspectives
β
βΌ
Tech Lead MAKES decision
β
βΌ
Communicate with rationale
β
βΌ
Team executes (even if disagreed)
β
βΌ
Evaluate together
Disagree and Commit
Team member disagrees:
Voice concerns clearly
β
βΌ
Tech Lead listens
β
βΌ
Decision still made
β
βΌ
Team member COMMITS to execution
β
βΌ
No sabotage, no "I told you so"
β
βΌ
Evaluate together later
π― 10. Checklist Tα»± ΔΓ‘nh GiΓ‘
Information Gathering
- CΓ³ Δủ ~70% thΓ΄ng tin cαΊ§n thiαΊΏt?
- ΔΓ£ hα»i Γ½ kiαΊΏn team members?
- ΔΓ£ hiα»u business context?
Analysis
- CΓ³ Γt nhαΊ₯t 2-3 options?
- ΔΓ£ list pros/cons của mα»i option?
- ΔΓ£ xΓ‘c Δα»nh trade-offs?
Decision Making
- QuyαΊΏt Δα»nh cΓ³ timely khΓ΄ng?
- CΓ³ clear rationale?
- ΔΓ£ communicate decision?
Execution & Learning
- CΓ³ execution plan?
- CΓ³ ngΖ°α»i own tα»«ng task?
- CΓ³ evaluation metrics?
- SαΊ΅n sΓ ng adjust nαΊΏu sai?
π‘ Tα»ng KαΊΏt
Tech Lead Decision Making:
1οΈβ£ Collect Info β 70% rule, khΓ΄ng cαΊ§n perfect
2οΈβ£ Analyze β Options + trade-offs
3οΈβ£ Decide β Timely, not perfect
4οΈβ£ Execute β Delegate & track
5οΈβ£ Learn β Evaluate & improve
Mindset:
Tech Lead mindset:
NOT perfect decisions
+
Timely decisions
+
Ownership
+
Learning
Key Principles:
βββββββββββββββββββββββββββββββββββββββββββ
β β
β 1. 70% information β decide β
β β
β 2. Good now > Perfect too late β
β β
β 3. Most decisions are reversible β
β β
β 4. Own the outcome (right or wrong) β
β β
β 5. Learn & iterate β
β β
βββββββββββββββββββββββββββββββββββββββββββ
βThe best decision is one thatβs made. The worst decision is one thatβs avoided.β
π TΓ i Liα»u Tham KhαΊ£o
- Book: Thinking, Fast and Slow β Daniel Kahneman
- Book: The Hard Thing About Hard Things β Ben Horowitz
- Article: Type 1 and Type 2 Decisions β Jeff Bezos
- Article: Architecture Decision Records β ADR templates
BΓ i liΓͺn quan:
- Tech Lead Time Management β CΓ‘ch PhΓ’n Bα» Thα»i Gian Hiα»u QuαΊ£ β 40% Coding, 25% Meetings, 20% Reviewβ¦
- How to Become a Tech Lead β Career Path, Mindset & Responsibilities β Career path vΓ 5 kα»Ή nΔng cα»t lΓ΅i.