Insights

Insights shows how Ogma is performing for this server. Use it to find the highest-leverage improvements to your content and process.

Time period

Plan	Available periods
Free	Last 7 days
Plus & Enterprise	Last 7, 30, or 90 days

Use the period selector at the top to change the window. All metrics and charts update for the selected range.

Key metrics

Metric	What it means	What to do about it
AI deflection rate	Share of closed tickets resolved by the bot without escalation	High and stable is good. Sudden drops often mean a content gap or a change in what users ask.
Top questions	The questions asked most often across tickets	If the same question keeps appearing and deflection is low, add or improve the matching doc or tip.
Tips learned	New tips added in the period	Review auto-learn output and remove bad tips in Tips or #learned-tips.
Knowledge gaps	Escalations where Ogma had no relevant content	Highest priority list for new docs — sort by frequency.
CSAT satisfaction	Helpful vs not helpful ratings	Low CSAT with high deflection can mean correct but unhelpful tone or completeness.

Charts

Insights includes visual breakdowns for the selected period:

• Ticket resolution — how closed tickets were resolved (bot resolved, escalated, other) — all plans
• CSAT breakdown — helpful vs not helpful share — all plans
• CSAT trend — daily rating counts over time — Plus & Enterprise only
• Performance by ticket type — closed, bot-resolved, and escalated counts grouped by type — Plus & Enterprise only

Charts need enough ticket volume to render; empty states explain when data is not yet available.

Knowledge gaps — take action

Each knowledge gap row links to the source ticket. From the gap list you can:

Action	Effect
Add to knowledge base	Opens a pre-filled Knowledge add sheet from ticket context
Create tip	Drafts a short tip from the conversation
Mark out of scope	Removes the gap from the list without adding content

Gaps are the fastest path from "Ogma escalated" to "doc exists for next time."

CSAT details

• Recent CSAT comments — written feedback after close (all plans)
• Deflected tickets — recently closed tickets the bot resolved without staff (Plus & Enterprise)
• Staff performance — per-staff closed tickets, claims, response/resolution times, message volume, and CSAT (Plus & Enterprise)

Staff performance

The Staff performance table attributes metrics to individual staff Discord accounts for the selected period:

Column	Meaning
Closed	Tickets where this staff member is recorded as the closer
Claimed	Tickets they claimed
Avg first response	Time from ticket open to their first public staff reply
Avg resolution	Time from open to close for tickets they closed
Messages	Staff replies sent (excluding internal notes)
CSAT rate	Share of positive CSAT on tickets they closed

CSAT is attributed to the closer in v1 — use ticket detail and transcripts when you need finer attribution.

Ask Ogma about support history (Enterprise)

Ask free-form questions about the selected period — "How many tickets involved refunds?", "What mood have customers been in?" — and get an evidence-led answer with links to the supporting tickets.

How it answers:

• Counting questions are computed over every ticket in the period, not a sample: each closed ticket gets a short topic analysis at close, and the question is matched against all of them by topic keywords plus semantic similarity. Counts are reported as lower bounds.
• Narrative evidence comes from the most relevant conversations, found by semantic search over per-ticket summaries embedded when tickets close. Open tickets are matched by their live content.
• Every claim in the answer must quote real ticket text; unsupported findings are discarded before you see them.

The first question you ask triggers a one-time background backfill that analyses and embeds closed tickets from the last 90 days, so coverage improves within minutes on first use. Customer identifiers are redacted before any ticket text reaches the model, and ticket summaries are never used to answer other customers' live tickets.

Staff can also ask from Discord with /ogma history — the answer is ephemeral (only the asking staff member sees it) and counts toward the plan's direct AI usage, like /ogma askgpt.

Using Insights with other pages

• Repeated question → Knowledge or Tips
• Gap cluster → write the missing doc, test in Playground
• CSAT drop on a type → review transcripts and Playground the same questions
• Many tips learned → review and remove bad ones in Tips or #learned-tips

Related

• Knowledge — source health and freshness
• Tickets — individual histories
• Settings — controls that influence deflection and tone
• Support ops — SLAs and quality flags (Plus/Enterprise)