What Quality is
Quality is a daily review of your team's conversations. Every day, AstraChat uses artificial intelligence to read each new conversation, scores it from 1 to 5, writes a summary of what the agent did well and what could improve, and flags low-scoring conversations for you to review.
The report also brings the day's work together on one page: who replied how much, how fast new customers got an answer, what was left open at the end of the day, what was closed without a reply and what was promised but not followed up.
Rating scale
| Score | Meaning |
|---|---|
| 5 · Excellent | Professional, empathetic service; the issue was fully resolved. |
| 4 · Good | The agent helped and resolved most of the issue. |
| 3 · Average | The agent was polite but did not fully resolve it. |
| 2 · Poor | Incorrect information or very little help. |
| 1 · Very poor | Rude, unhelpful, ignored the customer or never replied. |
Your customers see nothing. Quality only reads conversations. It sends no messages, changes no statuses or assignments and modifies nothing in AstraChat.
Who can use it and where to find it
A separate add-on. Quality is not included in any plan: it is contracted separately. Write to app@astranexo.com to turn it on for your account. It does not use your plan's AI credits.
Administrators only
Quality is available only to account administrators. An agent who tries to open it sees the message “Administrators only”.
How to open it
- Sign in to AstraChat with an administrator account.
- In the sidebar, open Reports.
- Click Quality, at the end of the list of reports.
The three tabs
- Reports: the reports already generated, with the date, conversations, how many were rated, how many were flagged, the average, the status and the trigger (Manual or Scheduled). Use View to open one. Below it is the list of Weekly and monthly summaries.
- Generate: creates the report for a specific day, or a weekly or monthly summary.
- Settings: business hours, reply target, threshold, and the sending of the daily report, the weekly summary and the monthly summary, with their recipients.
Generate a report
The easiest option is to turn on the automatic daily report (see Settings). You can also generate one by hand:
Step by step
- Open Reports → Quality and go to the Generate tab.
- In Report date, pick the day. Yesterday is preselected. Future dates are not allowed.
- To review only some channels, select them under Inboxes. Empty means all.
- Click Generate report. A bar shows the progress; when it finishes, click View report.
What is evaluated
- Conversations created that day are rated.
- The team workload sections cover every conversation with activity that day, even if it started earlier.
- Conversations where the customer never wrote (for example, notices your business sent) are not rated.
- There can be only one report per date. To redo it, delete the old one from the Reports tab with Delete and generate it again.
The report header
At the top of each report you see five figures and a bar showing how many conversations got each score:
Main figures
| Figure | What it means |
|---|---|
| Conversations | Conversations created that day that went into the report. |
| Rated | Those that got a score from 1 to 5. |
| Average score | The average of the rated ones. Green from 4, yellow from 3, red below. The 7-day average is shown next to it. |
| Flagged (≤N) | Those scoring at or below your threshold (3 by default). They carry the ⚑ mark. |
| Not applicable | Those not rated because they are “N/A” (see below). |
Team workload
This section summarizes each person's day. It covers every conversation with activity that day, not only new ones. WhatsApp groups are left out because they are usually the team talking among themselves.
Day totals
Active, New, Customer messages, Replies sent, Resolved and Open at end of day. Several figures show “7-day avg” underneath to compare with the previous week.
Table per person
| Column | What it shows |
|---|---|
| Messages | Replies the person sent to customers from AstraChat. |
| Conversations | How many conversations they replied in. |
| Resolved | How many conversations they closed. |
| Response (median) | Their typical reply time, counted within business hours if you set them. |
| First reply on target | Of the new conversations they answered first, how many were within the target (for example 8/10). |
| Open at end of day | Conversations assigned to them still open at midnight. |
| Closed unanswered | Conversations they closed without anyone replying to the customer. |
| Reopened | Conversations they closed that the customer came back to. |
| Quality | The average score of their conversations in this report, with the count in brackets. |
Replies from the phone
If someone answers a WhatsApp directly from the phone, AstraChat stores that reply in the conversation as a note marked 📱 Sent from WhatsApp. Because WhatsApp does not say who wrote it, these replies get their own row, From the phone. Answered WhatsApp calls also count as a reply.
Automated messages
Automatic notices and bot messages are counted separately, in a note under the table, and are not credited to anyone.
Tip: Ask the team to reply from AstraChat whenever they can. That way each reply is credited to whoever wrote it and the per-person table is fairer.
First reply to new conversations
Measures how long each new customer waited from their first message to the team's first reply, against your first-reply target (15 minutes by default). If you set business hours in Quality, only time within those hours counts.
Figures
- On target: the percentage answered in time. Green from 80%, yellow from 50%, red below.
- On time, Late and No reply: how many conversations fell in each group.
- Typical wait: the most common waiting time (the median).
- If you have more than one inbox, a table shows the same figures per inbox.
Tickets in the reports
If your account has Tickets turned on (Settings → Tickets), the daily report (in the email and on the page) and both summaries include a Tickets section.
What it shows
- New tickets and how many got a first reply, with the first reply (median).
- Solved, the time to solve (median) and reopened.
- Unresolved now, by status: New, Open, Waiting on customer and On hold.
- New by priority (Low, Normal, High, Urgent) and by type.
- A per-person table: Unresolved, Urgent or high unresolved, New and Solved.
- The most urgent unresolved tickets, with a link to each.
Two details. The first reply counts only replies sent from the ticket itself. “Unresolved now” is as of when the report was made.
Chats from Meta ads
If your business advertises on Facebook or Instagram with ads that open WhatsApp, this section gathers the chats that arrived through an ad that day. If no customer came from an ad, the section says so.
What it shows
- Chats from ads and how many were New customers (the rest were returning customers).
- The percentage answered within the target, how many got No reply and how many were Left open.
- Sales Won, Pending and Lost, based on the reading of each chat.
- A table per ad, another per assigned person, and the list of every chat with its first reply, its outcome and a link to the conversation.
Open at end of day
Lists the conversations still open at midnight, with the contact, the inbox, the person assigned now, how long it has been open and the reason:
Reasons
| Reason | What to do |
|---|---|
| Nobody replied to the customer | The most urgent: the customer never got a reply. Answer these first. |
| Customer waiting for a reply | They were answered before, but there are new unanswered messages. It shows how many and how long they have waited. |
| Replied; waiting on the customer or to be closed | The team already replied. If the case is done, resolve it; if you are waiting on the customer, use Snooze. |
| Snoozed | It is snoozed and will come back on its own when the customer writes or the date arrives. |
| No messages | An empty conversation; it can usually be resolved. |
Tip: To wait for the customer without leaving the conversation open, use Snooze instead of leaving it in Open. See Conversations.
Closed without a reply and customers who came back
Closed without anyone replying
Conversations resolved that day where the customer wrote and never got a reply. It shows who closed it, what the customer wrote and, if there is one, the internal note left when closing. Check whether the customer expected something.
Customers who came back after closing
Conversations the customer reopened 1 hour to 7 days after being resolved: the problem probably was not solved. They count for whoever closed them. Those that come back within an hour do not count, because they are almost always a “thank you”.
Leave a private note. If you served the customer another way (in the store, by phone call or from another number), say so in a private note before closing. The AI review reads those notes and does not penalize the missing reply in the chat.
Follow-ups promised yesterday
When an agent commits to write back to the customer (“I'll confirm tomorrow”, “I'll let you know when it arrives”), the review records the promise. The next day's report shows each promise as Followed up if someone wrote to the customer again that day, or No follow-up if not.
General phrases like “let me know if you need anything” do not count as promises. The section only appears when there were promises the day before.
What customers asked about, and sales
What customers asked about
A chart with the main reason for each rated conversation: Price or quote, Availability, Repair or order status, Technical support, Warranty or return, Complaint, Appointments, Purchase, Payments and invoices, General information or Other.
Sales
How many conversations ended in a Won sale (bought, paid, reserved or booked), how many are still Pending (showed interest and has not decided) and how many were Lost. Below, lost sales are listed with the reason, for example price, no stock or a slow reply, and a link to each conversation.
Messages by hour
A bar chart with customer messages and team replies in each hour of the day. If customer bars often outgrow reply bars at certain hours, that is where you need more people. Show as table gives you the same figures as numbers.
Quality of each conversation
At the end is the list of every conversation in the report, rated ones first. Each row shows the score, the contact, the inbox, the agent, the number of messages, a one-sentence summary with the feedback for the agent, and a button with the conversation number to open it.
Row states
- Score from 1 to 5: flagged rows (at or below the threshold) are highlighted with ⚑.
- N/A: there is no real customer to serve, for example bounce emails, spam, system notices, internal tests or conversations where the customer never wrote. They do not count toward the average or the flagged count.
- Error: that conversation could not be reviewed. The rest of the report completes normally.
7-day comparison: The header of this list, and several figures in the report, show “7-day avg”: the average of the previous seven days. Days with no activity at all (for example, days you are closed) are left out so they do not drag the average down.
Settings
Open Reports → Quality → Settings, change what you need and click Save settings.
Daily report
| Setting | What it does | Default |
|---|---|---|
| Send the daily report | On the ticked days, at the set time, generates the previous day's report and emails it. | Off |
| Time it is sent | What time the daily report is generated and sent. | 21:00 |
| Recipients | Email addresses that receive the report, separated by commas. Empty = no email is sent. | Empty |
| Days it is sent | Each report covers the day before; for example, untick Monday if you don't work on Sundays. With all days ticked, or none, it is sent every day. | All 7 days ticked |
Weekly summary
| Setting | What it does | Default |
|---|---|---|
| Send the weekly summary | Covers the 7 full days before the send day and compares them with the 7 days before that. Sent on a Monday, it covers Monday to Sunday. | Off |
| Day it is sent | The day of the week it is sent. | Monday |
| Time it is sent | What time it is sent. | 08:00 |
| Recipients | Empty = the same recipients as the daily report. | Empty |
Monthly summary
| Setting | What it does | Default |
|---|---|---|
| Send the monthly summary | Covers the previous calendar month and compares it with the month before. | Off |
| Day of the month | 1 to 28, so it exists in every month. | 1 |
| Time it is sent | What time it is sent. | 08:00 |
| Recipients | Empty = the same recipients as the daily report. | Empty |
For every report
| Setting | What it does | Default |
|---|---|---|
| Alert threshold | Conversations with this score or lower are flagged (≤ 1 to ≤ 5). | ≤ 3 |
| Time zone | Defines what “a day” is and which hour the data is shown in. | America/Costa_Rica |
| First reply target (minutes) | A new conversation meets the target if someone replies within this time (1 to 1440 minutes). | 15 |
| Business hours | Tick the days and hours your team works. Response times only count inside those hours. With no day ticked, all hours count. | None |
| Inboxes to evaluate | Which channels go into the automatic report. None selected = all. | All |
If the service was restarting at the exact send time, the report still goes out within the next 6 hours.
These hours do not affect your customers. Quality's business hours are for measuring only. Unlike an inbox's business hours (see Reports), they send no out-of-office messages.
Weekly and monthly summaries
One email per week or month with the totals of the period and the change against the previous period: ▲/▼ in % for counts, in points for the on-target percentage, and the score difference.
What it calculates
- It makes no new Claude ratings. The workload numbers (new conversations, customer messages, replies, resolved, first reply on target, typical first wait, new chats with no reply, closed without reply, reopened, open at the end of the period, and chats from ads) come from all conversations, every day.
- The quality score, flagged conversations, topics and sales come from the daily reports. If some days have no daily report, the summary says how many days the quality covers. Turn on the daily report and your summaries will be complete.
Sections of the summary, in order
- Para revisar (To review): the things to look at first.
- Las cifras (The numbers): a table with this period, the previous one and the change.
- Trabajo del equipo (Team work): per person, messages, resolved, first reply on target, median response (the median of each day's median), closed without reply, reopened, quality average with its count, low scores and sales won; plus the “Desde el teléfono” (From the phone) row.
- Día por día (day by day, weekly) or Semana por semana (week by week, monthly, plus the average customer messages per weekday).
- First reply by inbox.
- Tickets (if you have them turned on).
- Chats from Meta ads, per ad.
- Quality of service: the score distribution from 5★ to 1★, topics, sales, and the lowest-scored conversations and lost sales, each linking to the conversation.
- Customers waiting for a reply at the end of the period.
- Messages by hour.
Where to see them
Open Reports → Quality → Reports: the Weekly and monthly summaries list has View and Delete buttons.
Make one by hand
- Open the Generate tab and find the Generate a summary card.
- Choose Weekly (pick the last day of the week) or Monthly (pick the month). Only periods that have already ended can be chosen.
- Generate it: it is emailed to the configured recipients.
The report by email
If you enter addresses in the daily report's Recipients, each finished report is sent to them. The subject shows the date, the average score, the first-reply-on-target percentage and how many conversations were flagged, for example “[AstraChat] Daily report Nov 3, 2026 · 3.9/5 · 1st reply on target 43% · 7 flagged”.
The email follows this order: the summary numbers (with the 7-day average next to the average score), then To review, a View the full report button, the flagged conversations and finally the detail.
To review is a short list of what needs attention: first reply on target under 80%, new chats never answered, customers waiting at the end of the day, conversations closed without replying, follow-ups promised yesterday that nobody did, urgent or high tickets unresolved, new tickets without a first reply, customers who came back, flagged conversations and lost sales. If there is nothing, it shows a green “Nothing urgent” line.
The detail carries almost the same as the page: team workload, first replies, tickets, chats from ads, open at end of day, closed unanswered and reopened, promises, topics and sales, messages by hour and the flagged conversations, with links to each conversation. If a list is long, it shows the first ones and tells you to see the rest under Reports → Quality. If no conversation fell below the threshold, the email says so. Weekly and monthly summaries also arrive by email (see Weekly and monthly summaries).
How to use it with your team
Quality works best as a short, steady routine. We suggest this one:
Every morning (5 minutes)
- Read the email or open yesterday's report. Start with To review.
- Check Open at end of day, starting with “Nobody replied to the customer”, and assign each one.
- Check Closed without anyone replying and Follow-ups promised marked “No follow-up”. Write to those customers today.
Once a week (30 minutes with the team)
- Open the weekly summary (it arrives on the day you chose) and see what changed against the previous week, especially Para revisar and Trabajo del equipo. Once a month, do the same with the monthly summary.
- Compare the figures with the “7-day avg”: did the on-target percentage improve? Did open-at-end-of-day go down?
- Pick two or three flagged conversations and read them together. Talk about what could have been done differently, without looking for someone to blame.
- Also pick a conversation scored 5 and recognize whoever handled it.
- Review What customers asked about and Lost sales: if a question keeps coming up, create a canned response.
- Look at Messages by hour and adjust shifts if customers are waiting at certain hours.
- Agree on one improvement for the next week and check it at the next meeting.
Tip: The score is a guide, not a verdict. If you disagree with a rating, open the conversation and decide for yourself. To help the review understand the context, ask the team to leave private notes when something is solved outside the chat. You can create canned responses from Canned responses.
Frequently asked questions
Does Quality use my plan's AI credits?
No. Quality is a separate add-on and does not use Captain's AI credits.
Do agents see their scores?
Only administrators can open Quality. You decide what to share with the team, for example at the weekly meeting.
Why does a conversation say “N/A”?
Because there was no real customer to serve: a bounced email, spam, an automatic notice, a test or a message your business sent that the customer did not answer.
Can I redo a day's report?
Yes. Delete that date's report in the Reports tab and generate it again in Generate.
Do the weekly and monthly summaries use more AI?
They make no new ratings: they use the figures from all conversations and the scores from the daily reports. If daily reports are missing, the summary says how many days the quality covers.
How do I add or cancel it?
Write to app@astranexo.com.