Failure Diagnosis

Failure Diagnosis

When the system isn't working, diagnose which layer is failing.

Purpose

The optimization stack has three layers: Target, Algorithm, Friction. Intervening at the wrong layer wastes effort. This skill diagnoses the actual problem.

The Separation Principle

Check in order:

1. Is the target right? (Rare problem)
2. Is the algorithm right? (Sometimes the problem)
3. Is friction too high? (Usually the problem)

Most failures are friction failures. But check systematically.

Diagnostic Flowchart

START: Something isn't working
              |
              v
+--------------------------------------+
| Are you executing the behaviors?     |
| (Check compliance: >80%?)            |
+--------------------------------------+
              |
       +------+------+
       |             |
      YES            NO
       |             |
       v             v
+------------+   +-----------------------------+
| Check      |   | FRICTION FAILURE            |
| Algorithm  |   | Know what to do but not     |
| & Target   |   | doing it.                   |
|            |   | -> Run friction-audit       |
+------------+   +-----------------------------+
       |
       v
+--------------------------------------+
| Are outcomes moving in the right     |
| direction? (Check trends)            |
+--------------------------------------+
              |
       +------+------+
       |             |
      YES            NO
       |             |
       v             v
+------------+   +-----------------------------+
| WORKING    |   | Possible ALGORITHM FAILURE  |
| Keep going |   | Behaviors not producing     |
|            |   | outcomes.                   |
+------------+   +-----------------------------+
                          |
                          v
              +-----------------------+
              | Has enough time       |
              | elapsed?              |
              +-----------------------+
                          |
                   +------+------+
                   |             |
                  YES            NO
                   |             |
                   v             v
              +---------+   +---------+
              | WRONG   |   | WAIT    |
              | ALGO or |   | Lagging |
              | TARGET  |   | indica- |
              |         |   | tors    |
              | Review  |   | are     |
              | both    |   | slow    |
              +---------+   +---------+

Layer-Specific Diagnosis

Friction Layer (Most Common — Check First)

Symptoms:

• Low behavior compliance (<80%)
• "I know what to do but don't do it"
• Starting but not finishing
• Skipping logging/tracking
• Feeling of "too hard" or "too much"
• Procrastination, avoidance

Diagnostic questions:

• What specifically stops execution?
• Is getting started the problem? (Activation friction)
• Is continuing the problem? (Execution friction)
• Is tracking itself the friction?

Intervention: Run friction-audit skill

Algorithm Layer (Sometimes)

Symptoms:

• High behavior compliance (>80%)
• Outcomes not changing after sufficient time
• Effort without progress
• "I'm doing everything right but..."

Diagnostic questions:

• Are these behaviors actually connected to the outcome?
• Is the behavior being done correctly? (Quality not just quantity)
• Is there a better known approach for this target?
• Is something else interfering? (Confounding factors)
• Has enough time elapsed? (See timing table)

Intervention:

• Research evidence-based approaches
• Consult domain experts
• Re-run behavior-mapping with better algorithms

Target Layer (Rare but Critical)

Symptoms:

• Achieving metrics but not satisfied
• Proxy drift — number improves, actual goal doesn't
• Constant goal-changing, pivoting
• Targets conflict and paralysis results
• Success feels empty

Diagnostic questions:

• Is this metric measuring what you actually care about?
• If you hit this number, would you feel successful?
• Have priorities changed since setting this target?
• Are you optimizing a proxy while neglecting the real goal?

Intervention: Re-run target-definition skill

Expected Time to Results

Lagging indicators are slow. Before diagnosing algorithm failure, confirm enough time has passed:

Domain	Target Type	Minimum Time
Productivity	Output quality	2-4 weeks
Learning	Skill improvement	4-8 weeks
Health	Body composition	4-8 weeks
Finance	Savings growth	1-3 months
Habits	Automaticity	4-8 weeks
Writing	Quality improvement	4-8 weeks

If compliance is high but less than minimum time elapsed: Keep executing.

The Willpower Trap

When diagnosing friction failures, beware these responses:

• "I just need to try harder" — Rarely works
• "I need more discipline" — Willpower depletes
• "I'm being lazy" — Usually system problem, not character

Friction is structural. Willpower is a resource.

The answer is almost always: reduce friction, don't increase effort.

Output Format

diagnosis:
  date: [date]
  presenting_problem: "[user's description]"

  compliance_check:
    [behavior]: [%]  # Flag if <80%

  time_elapsed: [duration]
  expected_time_for_results: [duration]

  diagnosis: "[FRICTION|ALGORITHM|TARGET] FAILURE"
  layer: [friction|algorithm|target]

  reasoning: |
    [explanation of diagnosis]

  specific_issues:
    - "[issue 1]"
    - "[issue 2]"

  recommended_intervention: "[skill to run]"

  suggested_experiments:
    - "[experiment 1]"
    - "[experiment 2]"

Quick Reference

Compliance	Outcome Trend	Time	Diagnosis	Action
Low (<80%)	Any	Any	Friction	friction-audit
High (>80%)	Improving	Any	Working	Continue
High (>80%)	Flat	Short	Patience	Wait longer
High (>80%)	Flat	Long	Algorithm	Research alternatives
High (>80%)	Wrong direction	Any	Algorithm/Target	Deep review
Any	Success feels empty	Any	Target	target-definition