Appearance
Overview
The Analytics page answers two questions a support lead is asked every month: what is the queue actually about, and what is Overdesk doing for us. Both are drawn from your own workspace's numbers, not from an average.
Open Analytics in the sidebar, under Overview. The period selector in the top right sets the window every figure on the page covers, 7, 30 or 90 days, and each figure is compared with the equally long period before it. The choice is remembered for next time, and the same selector sits on Home and on Usage.
What Overdesk did for you
The first panel estimates how much writing Overdesk saved your team over the window, and shows the arithmetic rather than asking you to take it on trust.
| Figure | What it means |
|---|---|
| Drafts used | Drafts your team sent, either as written or after an edit |
| Rewritten by your team | Drafts that were replaced rather than used. These count as nothing saved |
| What it cost | What those drafts cost on your Anthropic key over the window |
| Per draft written | The same cost, divided by the number of drafts written in that window |
The time figure comes from comparing each draft with the reply your customer actually received. Overdesk measures how much of the draft survived, counts that many words, and divides by a typing speed. How time saved is calculated sets out the method in full, including what it deliberately leaves out.
The panel stays empty until conversations start closing: Overdesk can only compare a draft with a reply once the reply has been sent and the conversation graded.
What's happening in your queue
The second panel is built from triage. As Overdesk classifies conversations it records a product, an issue type and an urgency for each one, and this panel counts them over your window against the period before.
- By issue type and By product show which categories account for the queue, and how each one has moved. A category that appeared in the previous period and has since stopped is still listed, so you can see something going away as well as arriving.
- The urgency split counts how many conversations Overdesk called high, medium or low, each against the period before.
- Recent high-urgency conversations lists the most recent conversations triaged as high urgency, with their product and issue type.
Conversations are counted once each, by their most recent classification. A conversation that is reopened and classified again does not count twice, and a conversation Overdesk could not label still counts in the total but appears under no category.
This panel needs triage to be running. See Triage.
Needs attention now
The third panel lists the conversations that have been waiting longest, taken from your most recent morning briefing. The heading says how long ago that briefing ran, because the list is as fresh as the briefing rather than as fresh as the page.
Refresh generates a new briefing. It takes a minute or two, and the panel updates once it finishes.
When the morning briefing's schedule is off, the panel says so rather than suggesting briefings have stopped arriving: no new briefing comes on its own until you turn the schedule back on under Schedules. Refresh still runs one whenever you ask.
What needs a decision
The fourth panel lists the issues nobody has decided on yet, one row per issue however many analyses reported it, the 8 most severe first, with a line counting how many issues there are, how many are new this week and how many are getting worse. Each one opens on the Intelligence page, where your team can acknowledge, resolve or dismiss it.
Its stamp says when an analysis last listed its issues. It says nothing needs a decision only when no scheduled analysis is late, and either a scheduled analysis has listed its issues since it was last due or one listed them in the last 48 hours. An analysis is late when it was due more than 3 hours ago and has not run since, and the panel names it and when it was due. It also names any analysis whose last run failed or did not list its issues, because what that analysis would have found is missing.
The executive read
The last panel shows your most recent executive summary: a short written read of where the queue stands and what to do about it. It carries the same stamp saying when it was written, and Write a fresh one generates a new summary on demand.
These panels read what Overdesk has already written rather than writing something new each time you open the page, so opening Analytics never spends anything on your Anthropic key. Every report Overdesk has produced is kept on the Reports page.
What the page does not show
Analytics reads Overdesk's own records, so it covers conversations Overdesk has seen: drafted, triaged, or included in a report. It is not a live view of your help desk, and it is not a replacement for your help desk's own reporting.
Sentiment is not among the figures. Overdesk does not score how a customer feels, and a number that looked like one would be a guess. The nearest thing available is the frustration analysis, which you can run yourself from Analyses.