Part 013 — Story Points, T-Shirt Sizes, Unknowns, and Estimation Failure Modes

Positioning

Estimation bukan mesin prediksi presisi.

Estimation adalah mekanisme untuk:

• membangun shared understanding;
• membandingkan ukuran relatif;
• menemukan uncertainty;
• menemukan dependency;
• dan mendukung forecast.

Core thesis: nilai terbesar estimation bukan angkanya, tetapi percakapan yang membuat perbedaan asumsi terlihat.

1. What Estimation Is

Estimation menjawab:

• seberapa besar work relatif;
• seberapa kompleks;
• seberapa banyak unknown;
• seberapa besar coordination;
• dan seberapa besar delivery risk.

Estimation tidak menjawab dengan pasti:

• berapa jam persis;
• siapa yang paling produktif;
• atau kapan pasti selesai.

2. Effort, Complexity, and Risk

Effort

Jumlah pekerjaan.

Complexity

Jumlah hubungan, rule, dan state yang harus dipahami.

Risk

Probability dan impact dari outcome yang salah.

Uncertainty

Seberapa banyak informasi penting belum diketahui.

Dua story dengan effort serupa dapat memiliki estimasi berbeda karena uncertainty.

3. Relative Estimation

Relative estimation membandingkan item.

Reference story = 3 points

New story:
- similar scope
- more integration
- more unknown

Estimated as 5 or 8

Relative comparison biasanya lebih reliable daripada absolute hour prediction.

4. Story Points

Story points menggabungkan:

• effort;
• complexity;
• uncertainty;
• dan risk.

Story points bukan:

• hours;
• days;
• performance score;
• atau individual capacity.

Common scales

• Fibonacci-like: 1, 2, 3, 5, 8, 13
• powers: 1, 2, 4, 8
• custom relative scale

Scale matters less than consistent meaning.

5. T-Shirt Sizes

Useful untuk early-stage planning.

• XS
• S
• M
• L
• XL

Benefits:

• avoids false precision;
• fast;
• good for roadmap-level comparison.

Limitations:

• too coarse for Sprint planning;
• can hide large uncertainty.

6. Ideal Day Anti-Pattern

Ideal day assumes uninterrupted work.

Reality includes:

• review;
• meetings;
• waiting;
• debugging;
• CI;
• QA;
• and coordination.

Converting points to ideal days often reintroduces false precision.

7. Unknowns

Known known

Team understands.

Known unknown

Question identified.

Unknown unknown

Important uncertainty not yet visible.

Estimation divergence often signals hidden unknown.

8. Estimation Conversation

Ask:

What makes this larger?
What can go wrong?
What dependency exists?
What evidence is missing?
What is included in Done?
What assumptions differ?

9. Reference Stories

Reference stories improve consistency.

Choose examples:

• small;
• medium;
• large;
• well understood;
• completed recently.

Avoid outdated references when Definition of Done changed.

10. Estimation and Definition of Done

Estimate must include all work needed for Done:

• implementation;
• review;
• testing;
• integration;
• documentation;
• observability;
• migration;
• release readiness.

If team estimates only coding, forecast is systematically wrong.

11. Estimation and Story Slicing

Large estimate may mean:

• story too broad;
• too many rules;
• too many dependencies;
• or insufficient refinement.

Estimation should trigger slicing, not only assign a larger number.

12. Estimation and Risk

A risky item may be ordered earlier even if effort small.

Estimate should expose risk, but not hide risk inside a number.

Record risk separately.

13. Planning Poker Preview

Planning poker combines independent estimates and discussion.

Value comes from divergence.

If everyone immediately agrees, verify whether assumptions are genuinely shared.

14. Senior Engineer Influence Risk

Senior engineer estimate can anchor the team.

Safer pattern:

• estimate independently;
• reveal simultaneously;
• ask lower estimates first;
• ask higher estimates;
• discuss assumptions.

15. Estimation Failure Modes

Point-to-hour conversion

Creates false precision.

Individual estimation

Ignores team delivery.

Estimate as commitment

Punishes learning.

Velocity target

Encourages gaming.

Estimate before understanding

Produces meaningless numbers.

Endless debate

Estimation cost exceeds value.

Senior anchoring

Others follow authority.

Hidden work

Testing and release omitted.

16. Process Smells

• every story is 3 points;
• large story accepted without split;
• estimates changed by manager;
• points compared across teams;
• completed points used for performance;
• estimation takes longer than refinement;
• and unestimated urgent work dominates Sprint.

17. Worked Example: Approval Delegation

Story appears small:

Allow approver delegation.

Hidden complexity:

• delegation period;
• circular delegation;
• audit;
• revocation;
• timezone;
• existing approval task;
• authorization;
• compatibility.

A divergence from 3 to 13 points indicates different mental models.

Discussion is the value.

18. Senior Engineer Operating Model

• expose hidden scope;
• avoid anchoring;
• explain technical risk;
• include operational work;
• suggest spike when uncertainty dominates;
• and challenge false precision.

19. Internal Verification Checklist

• What estimation scale is used?
• Are points converted to hours?
• Are estimates team-owned?
• What reference stories exist?
• Is Definition of Done included?
• How are spikes estimated?
• Are bugs estimated?
• Is velocity used for performance?
• How often estimation differs significantly from actual flow?

20. Practical Exercises

Exercise 1

Select five completed stories and compare:

• estimate;
• uncertainty;
• actual blockers;
• missed work.

Exercise 2

Create reference stories for small, medium, and large.

Exercise 3

Take one large item and split it until each slice is estimable.

Exercise 4

List what your team includes and excludes from estimation.

21. Part Completion Checklist

You are done if you can:

• separate effort, complexity, risk, and uncertainty;
• use relative estimation;
• explain story points correctly;
• use T-shirt sizing appropriately;
• detect false precision;
• facilitate divergence discussion;
• and prevent estimation from becoming performance measurement.

22. Key Takeaways

1. Estimation is a conversation.
2. Story points are relative.
3. Points are not hours.
4. Divergence reveals assumptions.
5. Definition of Done affects estimate.
6. Large estimates should trigger slicing.
7. Risk should remain explicit.
8. Senior engineers must avoid anchoring.
9. Estimates support forecast, not certainty.
10. Internal conventions must be verified.

23. References

Conceptual baseline:

• The Scrum Guide.
• Agile estimation and relative sizing practices.
• General forecasting and uncertainty-management practices.

These concepts do not describe internal CSG processes.