Jede Vermutung wird mit einer Prüfung verknüpft, die möglichst viel Information pro Test liefert. Dadurch entsteht eine Reihenfolge, in der die stärksten Hypothesen zuerst unter Druck geraten und die Diagnostik nicht im Kreis läuft. Der Ansatz verbindet Debugging mit sauberer Evidenz.

1. 1

   Symptom

   Symptom view

   

   Goal

   Symptom view: Operationalise observed failure and collect Is/Is-Not cases.

   Inputs for this step

   • Set symptom, affected/unaffected cases, current signals, and safe test boundaries; separate containment from diagnosis.
   • Precise symptom, architecture and changes, hypothesis board, predictions, safe tests, and evidence log

   Actions

   • Operationalise observed failure and collect Is/Is-Not cases.

   Output

   Symptom view

   Operationalise observed failure and collect Is/Is-Not cases.

   Quality check

   Is success/failure measurable and free of causal claims?

   Pitfall

   Database problem is written as symptom. State only observable behaviour, scope, and time.

   Timing: 10–20 minutes

2. 2

   Hypotheses

   Hypothesis space

   

   Goal

   Hypothesis space: Derive several mechanistic explanations from signals, architecture, and changes.

   Inputs for this step

   • Symptom view

   Actions

   • Derive several mechanistic explanations from signals, architecture, and changes.

   Output

   Hypothesis space

   Derive several mechanistic explanations from signals, architecture, and changes.

   Quality check

   Does each explain affected and unaffected cases?

   Pitfall

   Team pursues only first plausible idea. Add a competing explanation and null hypothesis.

   Timing: 10–20 minutes

3. 3

   Predictions

   Prediction matrix

   

   Goal

   Prediction matrix: Write observable predictions and falsifying signals for every hypothesis.

   Inputs for this step

   • Hypothesis space

   Actions

   • Write observable predictions and falsifying signals for every hypothesis.

   Output

   Prediction matrix

   Write observable predictions and falsifying signals for every hypothesis.

   Quality check

   Are predictions recorded before testing and discriminating?

   Pitfall

   Tests are reinterpreted after results. Record expected findings and exclusion rule before execution.

   Timing: 10–20 minutes

4. 4

   Test

   Test result

   

   Goal

   Test result: Run safest highest-information test, changing one condition at a time.

   Inputs for this step

   • Prediction matrix

   Actions

   • Run safest highest-information test, changing one condition at a time.

   Output

   Test result

   Run safest highest-information test, changing one condition at a time.

   Quality check

   Does test vary exactly the discriminating condition and remain reversible?

   Pitfall

   Multiple restarts destroy diagnostic information. One change per test; label emergency action separately.

   Timing: 15–90 minutes

5. 5

   Decide

   Diagnostic finding

   

   Goal

   Diagnostic finding: Update hypotheses against all evidence, record mechanism, and hand over next action/residual uncertainty.

   Inputs for this step

   • Test result

   Actions

   • Update hypotheses against all evidence, record mechanism, and hand over next action/residual uncertainty.

   Output

   Diagnostic finding

   Update hypotheses against all evidence, record mechanism, and hand over next action/residual uncertainty.

   Quality check

   Does leading hypothesis explain all signals with a serious alternative tested?

   Pitfall

   One successful test is called full root cause. State diagnostic boundary and pass remaining cause question separately.

   Timing: 15–30 minutes