Three numbers sit at the top of Insights, with a plain-English sentence above them summarising the period. They are the closest thing to a headline your support has.
The three headlines
Each shows the figure for the period you have selected, and how it has changed against the previous period of the same length.
On a new workspace they say First period with data instead of a change, because there is nothing to compare against yet.
Average first response
How long, on average, a customer waited for the first reply.
This is the number customers feel most, and the one most worth improving. It is also the easiest to flatter: if Chloe answers instantly on chat, this number becomes excellent without your team getting any faster. That is a real improvement for customers. Be clear about what it is measuring.
Average resolution time
How long from a conversation starting to it being resolved.
Averages hide their shape, and this one hides it more than most. A handful of threads that dragged on for weeks will drown out fifty that were sorted in ten minutes.
So treat a rising resolution time as a prompt to go and find the slow threads rather than a verdict on the team. Overdue is the better daily tool for exactly that reason: it is about individual conversations, not an average.
Customer satisfaction
The average star rating customers gave, out of five, from the post-chat survey.
It is the only number here that comes from customers rather than from timestamps, which makes it the most valuable and the most easily misread.
The star breakdown
Beside the average is the distribution. It shows how many people gave five stars, four, and so on, with a percentage for each row.
Read this before the average, every time. An average of 3.7 from two five-star ratings and one one-star is not a middling performance; it is two delighted customers and one upset one, and those need different responses.
Each star row opens its threads, so a one-star row is a two-click route to exactly what went wrong.
How many people actually rate you
Under the breakdown is the response rate. It is the share of resolved threads that asked for a rating and actually got one. In practice it is low, often under ten per cent.
When the count is small, a Low sample tag appears next to it. Take that seriously. Three ratings is not a measurement, and moving from 3.7 to 4.2 on that basis tells you nothing.
Threads by status
Below the headlines, every conversation from the period is sorted by where it stands right now: Open, Waiting on customer, Snoozed, Resolved. A line underneath adds how many were closed without needing a reply, and how many were marked spam.
The important word is "now". This is not what happened during the period; it is where those conversations have ended up. A large Open figure on a period that finished a fortnight ago is a backlog, not a snapshot.
Kudos
At the bottom is Kudos. These are the actual words customers wrote alongside five-star ratings, with the teammate who earned them.
It only shows five-star ratings that came with a comment, so an empty section usually means nobody wrote anything, not that nobody was pleased. Widening the date range normally finds some.