Improve · Analytics
Numbers you can open.
A report that cannot show you the conversations behind a number is a rumour. Every figure in BlendDesk — volume, first response, resolution, channel mix, agent, Scout — drills back to the threads it came from.
Illustrative demo with fictional data.
What you can see
Four reports, and nothing you have to assemble by hand.
Pick a report below. The figures are fictional; the layout and the metrics are the product.
- Volume
- Conversations, messages received and sent, by day, week, month or year, with business hours on or off.
- First response
- How long a customer waited for the first human or Scout reply, as an average over the range and day by day.
- Resolution
- Resolution time and resolution count, so a fast first reply that fixes nothing has nowhere to hide.
- Channel mix
- Where the volume actually arrives — WhatsApp, Instagram, Messenger, email, web chat — before you staff for it.
- Agent breakdown
- The same six metrics per agent, per team, per tag and per inbox, one breakdown at a time.
- Scout breakdown
- What the agent handled, what it handed over, and what is waiting on a person.
Scout
Handled
162
Resolved without a person
69%
Handoff rate
31%
Average response
4.2 s
Knowledge gaps
11
Approvals
7
2 pending
Guardrail actions
5
Daily volume
Last 7 daysOutcomes
Scenario frequency
| Metric | Example value |
|---|---|
| Handled | 162 |
| Resolved without a person | 69% |
| Handoff rate | 31% |
| Average response | 4.2 s |
| Knowledge gaps | 11 |
| Approvals | 7 (2 pending) |
| Guardrail actions | 5 |
| Day | Conversations |
|---|---|
| Sunday | 21 |
| Monday | 26 |
| Tuesday | 24 |
| Wednesday | 31 |
| Thursday | 33 |
| Friday | 15 |
| Saturday | 12 |
| Outcome | Conversations |
|---|---|
| Answered | 112 |
| Handed over | 50 |
| Scenario | Conversations |
|---|---|
| Order status | 58 |
| Returns | 31 |
| Delivery times | 27 |
| Product questions | 22 |
| Other | 24 |
Fictional data. Illustrative demo with fictional data. Example numbers for illustration only.
Scout's own numbers
Resolved without a person, handoff rate, approvals pending.
Three numbers decide whether an AI agent is earning its place, and the third is the one vendors leave out. If approvals are piling up, Scout is not saving anyone time — it is queueing work.
What the report tracks
- Resolved without a person
- Conversations Scout closed where no agent replied afterwards. Counted once per conversation.
- Handoff rate
- How often Scout stopped and brought in a teammate, split by the reason it stopped.
- Approvals pending
- Actions waiting for a decision, and how long they have been waiting.
- Knowledge gaps
- Questions that retrieved nothing. This is your content backlog, ordered by frequency.
- Guardrail actions
- Which guardrail fired, what it did, and on which conversations.
- Scenario frequency
- Which declared scenarios customers actually hit, and which you wrote for nobody.
- Tool failures
- Calls that errored or timed out, with the endpoint and the run.
Scout figures appear once Scout is live in at least one inbox.
How Scout worksScout
Handled
162
Resolved without a person
69%
Handoff rate
31%
Daily volume
Last 7 daysOutcomes
Scout figures as tables
| Metric | Example value |
|---|---|
| Handled | 162 |
| Resolved without a person | 69% |
| Handoff rate | 31% |
| Average response | 4.2 s |
| Knowledge gaps | 11 |
| Approvals | 7 (2 pending) |
| Guardrail actions | 5 |
| Day | Conversations |
|---|---|
| Sunday | 21 |
| Monday | 26 |
| Tuesday | 24 |
| Wednesday | 31 |
| Thursday | 33 |
| Friday | 15 |
| Saturday | 12 |
| Outcome | Conversations |
|---|---|
| Answered | 112 |
| Handed over | 50 |
| Scenario | Conversations |
|---|---|
| Order status | 58 |
| Returns | 31 |
| Delivery times | 27 |
| Product questions | 22 |
| Other | 24 |
Fictional data. Illustrative demo with fictional data. Example numbers for illustration only.
Drill-down
Every row opens the conversations behind it.
A satisfaction score is an opinion until you read the conversations that dragged it down. Each metric, rating, tag and guardrail action is a filter over real threads, and opening it takes you to them in the inbox with the filter already applied.
- From a number to a thread
- Open the conversations behind any figure in one click — same filters, same workspace your team already uses.
- Comments stay attached
- A rating keeps its comment, its contact and the agent who handled it, so feedback is never an anonymous bar on a chart.
- Export when you need to
- Any report downloads as CSV with the same rows, for the spreadsheet somebody insists on.
Ratings are collected per inbox. On WhatsApp the rating request is a message template, so Meta approves it before it can send.
Customer ratings
Responses
186
Satisfaction
83%
Response rate
42%
Distribution
| Contact | Agent | Rating | Comment |
|---|---|---|---|
| Mariam Ahmed | Nour Hassan | Excellent | Quick answer about my delivery date, thank you. |
| Karim Adel | Omar Khaled | Good | Helpful, though I had to explain my order twice. |
| Salma Fathy | Dina Sameh | Excellent | My exchange was sorted in one chat. |
| Hana Mostafa | Nour Hassan | Average | The reply took a little while this morning. |
Fictional data. Illustrative demo with fictional data.
Analytics is not evaluation
Analytics tells you what happened. Evaluations tell you whether a change made it better.
A dashboard is a rear-view mirror: traffic moved, a number went up, and a dozen things changed that week. That is not evidence that your new scenario helped.
- Analytics
- Observational. Live traffic, every conversation, no control. Good for spotting the problem and staffing for it.
- Evaluations
- Experimental. Replay the same past conversations against the old and the new configuration and compare run to run.
- Together
- Analytics finds the gap; an evaluation proves the fix; the trace shows what the fix did on a specific conversation.
Next
Keep reading
- Evaluations and tracesScore a change against real past conversations before it ships.
- ScoutThe agent whose numbers appear in that report.
- Guardrails and approvalsWhat the approvals-pending figure is actually queueing.
- InboxWhere every drill-down lands.
- AutomationsRules fire too, and every firing is on the conversation.
Questions
Questions about reports.
Which reports are there?
An overview with live queue counts and traffic heatmaps, a conversations report with six metrics, breakdowns by agent, team, tag and inbox, customer ratings, SLA tracking, and Scout's own report.
Can I see the conversations behind a number?
Yes — that is the point of the page. Every metric, rating, tag and guardrail action opens the matching conversations in the inbox with the filter applied.
Can I export?
Yes. Report pages download as CSV with the same rows you see on screen.
Are the numbers on this page real?
No. Every figure in the examples belongs to Nile Home, a store that does not exist. They show how the reports are laid out, not results to expect.
How is a Scout resolution counted?
A conversation Scout handled where no agent replied afterwards, counted once per conversation. If a person steps in later, it stops counting as resolved without a person.
Does a dashboard tell me whether my change worked?
No, and we would rather say so. Live traffic is uncontrolled. To know whether a change helped, replay past conversations against both configurations in an evaluation and compare the runs.
Bring last month's numbers.
We will show you the same questions answered in BlendDesk, and where each figure opens the conversations underneath it.