Estimation Anti-Patterns
Purpose: Use this reference when Harvest includes estimated effort and you need to explain limitations or avoid misleading productivity claims.
Contents
- LOC pitfalls
- Estimation anti-patterns
- Fatigue warnings
- Harvest-specific guardrails
LOC Pitfalls
| ID | Problem | Why it matters |
|---|---|---|
EA-01 |
Solution complexity != problem complexity | Hard work can produce little code |
EA-02 |
Language dependence | LOC varies by stack |
EA-03 |
Refactoring penalty | Code reduction looks falsely "unproductive" |
EA-04 |
Non-coding work ignored | Design, review, debugging are invisible |
EA-05 |
No quality correlation | More code can mean more defects |
EA-06 |
Copy-paste incentive | Volume can displace maintainability |
EA-07 |
Generated code pollution | Boilerplate distorts effort |
EA-08 |
Team contribution invisibility | Mentoring and review disappear |
Estimation Anti-Patterns
| ID | Anti-pattern | Guardrail |
|---|---|---|
EA-09 |
Optimism bias | Prefer historical data and buffers |
EA-10 |
Anchoring | Re-estimate when new information arrives |
EA-11 |
Parkinson's law | Separate buffer from expected effort |
EA-12 |
Ignoring scope creep | Version estimates when scope changes |
EA-13 |
Using the fastest person as baseline | Report team-safe ranges |
EA-14 |
Assuming linear effort | Apply complexity multipliers in larger or riskier work |
Fatigue Warning
Long-hours metrics need explicit caution:
- Sustainable work beats raw hours.
- Night/weekend work
> 10%should trigger a health warning. - Do not reward long hours as productivity.
Harvest Guardrails
- Mark hours as estimated.
- Prefer
min / expected / maxwhen reporting to managers or clients. - Exclude auto-generated files when they materially skew effort.
- If estimate accuracy falls below
60%, recommend reviewing the estimation process. - AI-assisted coding (Copilot, Claude Code, Cursor, etc.) decouples LOC from human effort. When AI usage is high, prefer review-cycle and pickup-time signals over LOC-derived hours and document the AI posture in the report.