[object Object]

From Incident to Problem

Recurring or high-impact incidents convert to problem records. The conversion is one click in the agent UI (Actions -> Convert to Problem) and the resulting problem inherits the parent incident as a related record. Multiple incidents can link to a single problem so that resolution prevents the next 14 reports of the same issue. Configure the conversion criteria via Workflow Automator: any P1 incident, any incident with three or more linked duplicates in 30 days, or any incident lasting more than the SLA-resolution target.

// Convert an incident to a problem via API
fetch(`https://${domain}.freshservice.com/api/v2/problems`, {
  method: 'POST',
  headers: { Authorization: auth, 'Content-Type': 'application/json' },
  body: JSON.stringify({
    subject: 'Recurring DB connection pool exhaustion',
    description: 'See linked incidents 12345, 12390, 12421',
    requester_id: 12,
    priority: 3,
    status: 1,
    associate_ci: ['db-prod-01'],
    custom_fields: { incident_count_30d: 14 }
  })
});

RCA Frameworks

Pick one framework per investigation and stick with it. The 5 Whys works for linear cause chains and is fastest. Ishikawa (fishbone) works when contributing factors come from multiple categories — people, process, technology, environment. Kepner-Tregoe works for hard-to-reproduce intermittent issues where you need to compare what is and isn’t affected. Document the chosen framework in the problem record so future readers see the reasoning. Freshservice’s problem template ships sections for each framework — pick one and delete the others before saving.

Known Errors

Problems with documented workaround and pending permanent fix become Known Errors. Publish the workaround to the agent-facing KB so the service desk resolves similar incidents in 2 minutes instead of 30. Tag KB articles with known_error and link to the problem record so the relationship is bidirectional. The Known Error Database (KEDB) inside Freshservice is searchable from the incident UI — agents see “this matches Known Error X” suggestions automatically.

Reducing Incident Volume

Tie problem work to incident reduction. After a problem closes, run a 30-day and 90-day report comparing incident volume on the affected CI before and after. Quantifies prevention value in dollars when paired with average incident cost. The metric that matters is “incidents avoided since problem resolution” because it justifies the investment in problem management as a discipline rather than as overhead. A problem that prevents 50 incidents at $200 each pays for the analyst time many times over.

CAB Integration

Problems drive changes. The fix for a problem is usually a change — code patch, config update, infrastructure rebuild. Link the change record to the problem so the audit trail flows: incident -> problem -> RFC -> change -> validated fix. Without the link, problem management becomes paperwork that nobody reads. Configure the workflow so closing a problem requires either a linked change or an explicit “no fix required” justification.

Common Failure Modes

Three. First, problems opened for every incident regardless of recurrence — produces a backlog that never gets investigated. Reserve problems for genuine root-cause work; not every incident needs RCA. Second, RCA written by the on-call who fixed the incident; they’re too close to it and the analysis becomes self-justification. Rotate RCA authorship to a peer. Third, no closure date on problems — problems sit open for years with stale workarounds; require a target close date on creation and review weekly.

What Changed in 2026

Freshservice added Freddy-assisted RCA suggestions: open a problem, click “Suggest possible causes,” and Freddy returns a ranked list based on similar problems in your tenant and across anonymized aggregate data. Treat as a starting point for human review, not as a verdict. The accuracy is good enough to surface non-obvious correlations (a config change three weeks earlier that nobody linked to the current symptom) but not good enough to skip human judgment.

What to do this week

Pull the top 10 incidents by recurrence count over the last 90 days. Each one is a candidate problem. Open the highest-impact two as problems and assign owners with a 30-day target close date.

[object Object]
Share