What to include in a network RCA
| Field | Record this |
|---|---|
| Incident and scope | Incident ID, owner, severity, affected services, sites and incident window with timezone. |
| Impact | Observed user symptoms, affected flows and known limits of the impact assessment. |
| Timeline | Detection, investigation, decisions, approvals, mitigation and restoration. |
| Evidence register | Device or system, collection time, command or query, relevant result and access gaps. |
| Cause and confidence | Supported causal explanation; separate contributing conditions and untested hypotheses. |
| Response | Actions taken, approver, change reference and outcome. |
| Verification | The tests that demonstrate service recovery and any remaining issues. |
| Prevention | Follow-up action, accountable owner, target date and completion evidence. |
Distinguish the symptom, trigger and root cause
“The application timed out” describes a symptom. “A route changed” describes a possible trigger. Neither statement alone explains why the service became unavailable. Connect the observed failure to the forwarding or policy behavior that affected the actual flow.
Keep a small evidence register. For every conclusion, record the device or tool, the time of collection and a reference to the relevant output. Align timezones and identify clock uncertainty before building a timeline. If only one side of a connection is visible, state that limitation.
Avoid rewriting the timeline to fit the first explanation. Record alternative hypotheses and the observation that ruled each one out. If the evidence is insufficient, publish a provisional finding with the next verification step instead of declaring a root cause.
Worked example: application timeout after a policy change
Illustrative example, not a customer incident: a branch loses access to an application after a firewall policy update. The initial report is a timeout. The investigation records the application destination and port, checks recent changes and finds deny events matching the affected flow during the incident window.
A reviewed configuration comparison shows that the intended allow rule is no longer effective for that flow. The team restores the approved policy state through its change process. Repeat application transactions from the affected branch succeed, and the corresponding traffic is observed as permitted. These observations support the policy explanation more strongly than the timing alone.
The RCA should record the actual rule evaluation evidence and the authorized restoration, not simply say “firewall issue.” Potential prevention includes a relevant policy test before a future change and an application-level check afterward. Assign an owner and due date to each agreed action.
Where NetPrompt AI can help
NetPrompt AI’s alert investigation workflow can gather evidence, draft root-cause analysis and update or create a matching incident through configured integrations. That can reduce repeated copying between monitoring, device sessions and ticketing tools. The operational team still needs to review whether the evidence supports the conclusion.
This downloadable template is a suggested reporting structure, not a claim that every field is populated automatically by default. Map the required fields and evidence references to your workflow and ITSM configuration. Capture unavailable device access explicitly and keep approval and verification records with the incident.
Review before closing the analysis
- Can a reviewer trace every material conclusion to an observation?
- Are the reported impact and restoration time supported by tests or monitoring?
- Are temporary mitigation and permanent correction clearly distinguished?
- Are remaining unknowns assigned a follow-up rather than omitted?
- Do corrective actions have an owner, due date and measurable completion condition?
Common questions
- Is an RCA the same as an incident ticket?
- An incident ticket coordinates response and restoration. An RCA explains the causes and follow-up actions. They can be linked or stored together, but closing a ticket does not by itself complete the analysis.
- Should an AI-generated RCA be reviewed?
- Yes. Check the underlying evidence, the incident scope and any proposed action. Fluent text does not establish that a causal explanation is correct.
- Can I use the template without NetPrompt AI?
- Yes. Download the Markdown file and adapt it to your own incident process. It is an ungated reference template and does not require an account.
Continue exploring
Further technical reading
External references explain general practices; they are not endorsements of NetPrompt AI.