Insights & reliability
The home page of your dashboard. A single screen that tells you whether the week is going well, and where to look if it isn't.
What it shows
Insights aggregates every conversation analyzed in your selected window into a one-page operational view: how much volume came in, how fast you replied, how good the replies were, and how customers felt about it.
It is the only module you need to open daily. Everything else is either a drill-down or a longer-cycle review.
The KPI strip
Four headline numbers, updated every time a new conversation closes:
- Conversation volume — count over the window, with the period-over-period delta. A spike here is usually the first signal of an incident or a release with a regression.
- First response time (FRT) — median and p90. p90 is the one to watch — the median can stay flat while the long tail explodes.
- Average quality score — mean of all reply scores in the window, 0 to 100. Below 70 is the line where customers start to notice.
- Negative sentiment rate — share of conversations where the final customer sentiment was negative. Lagging indicator of trouble.
Tip
The reliability tile
Tells you how much you can trust the AI quality score. It's the correlation between scores ResponZ assigns and CSAT ratings your customers leave on the same conversations.
- Above 0.7 — strong. Treat the AI score as a trustworthy proxy for satisfaction when CSAT is missing (and CSAT is usually missing for 70%+ of conversations).
- 0.5 – 0.7 — usable but calibrate. Spend a few minutes weekly in the Quality module confirming or correcting low scores.
- Below 0.5— the AI isn't tuned to your context. Review your scoring policy in Settings (tone, custom rules, sensitive topics).
Top demanding customers
Ranks accounts by conversation volume in the window. Useful to spot the 1–2 accounts that absorb a disproportionate share of your team's time — often a signal for a CSM conversation, not a support fix.
Workflows
Daily standup
Open Insights, eyeball the four KPIs and the deltas. If a number is red, jump to its source module (Quality for score, Top questions for volume, Targets for SLA) and decide before the standup.
Weekly review
Compare the current 7 days with the previous 7. Capture in a doc: what moved, why (look at Trends for releases), and one action for the next week. Five minutes max.
Settings
The window selector (top right) controls all KPIs and the reliability calculation. The default is 30 days. Changing it does not consume credits — it's a view over data already analyzed.