Failure Analysis
The specific reasons conversations are failing, so you know exactly what to fix.
Shown in Companion Insights only — this card isn't currently available on the org-wide Dashboard.
The Failure Analysis card is a searchable, sortable, filterable list of every detected failure, grouped by root cause and severity — not just a count of how often things go wrong, but why.
What you see
| Column | What it shows |
|---|---|
| Failure Reason | One of: knowledge gap, hallucination, misunderstood intent, incorrect info, missing product knowledge, capability gap, tone mismatch, policy block, unclear question, out of scope, or technical error. |
| Severity | Critical, High, Medium, Low, or None. |
| Count | How many conversations hit this failure. |
Hover any row for a fix recommendation specific to that failure reason.
How to use this
- Sort by Severity first — Critical and High entries deserve attention before anything else, regardless of count.
- A high count of Knowledge Gap or Missing Product Knowledge failures means it's time to add or update training resources.
- Hallucination or Incorrect Information failures are the most urgent to fix — they erode trust fastest. Trace them back to specific source material and correct it.
- Use this to explain movement in Resolution Rate and Response Quality — a dip in either usually shows up here as a spike in one specific failure reason.
Last updated on