Root cause

When should I use this?

• You need a defensible explanation to share with your team or leadership.
• You want to see why AutoObserve reached a conclusion — not just the conclusion.
• You are ready to act on a recommended next step with inspectable backing.

Requirements

• Investigation with tested hypotheses and correlated evidenceRequired
• Sufficient signal coverage for the incident windowRequired
• Team alignment on recommended action scopeOptional

Steps

1. Read the current best explanation

The explanation card includes:

• Why — supporting reasons linked to evidence
• Uncertainty — gaps that reduce confidence (e.g. missing profiling data)
• Alternative explanations — weaker hypotheses still on record
• Impact — blast radius across dependent services
• Recommended next step — e.g. review or rollback a deployment

RCA is the output of investigation — not the product category. AutoObserve maintains the strongest explanation supported by evidence available right now.

2. Verify the causal chain

Trace the chain from change to symptom:

Deployment → Configuration change → Saturation → Queuing → Latency

Reject explanations where downstream symptoms are treated as root cause, or where temporal order contradicts the hypothesis.

3. Inspect the audit trail

Open the full reasoning audit trail to see how evidence, hypothesis tests, and confidence updates led to the explanation. Intelligence is traceable back to source evidence.

Optional deep dive: Compare the checkout-latency explanation on Demo against your own investigation.

Did it work?

• Explanation lists inspectable why-reasons
• Uncertainty and alternatives are visible
• Recommended next step is actionable
• Ready to run your first live investigation

Troubleshooting