Service Metrics
Sidebar: Communication → Metrics
Module: Inbox — the screen appears with the Inbox module.
Permission: Service metrics — managers and analysts read it by default, members do not. Revenue figures also need permission to view Financial. See Settings → Permissions.
The Inbox is where the team answers. Metrics is where its manager reads how that work is going: how long customers wait, how much the team carries and when, how much of it the AI and the automations already take, what slips through, and what each attendant’s conversations produce.
Every figure is computed from what the Inbox already records — the conversations, their messages, who wrote each one and who resolved them. Nothing is typed in or estimated.
The filters
One row at the top scopes the whole screen.
| Filter | What it does |
|---|---|
| Start / End | The period, to the second. It opens on the current month: the 1st at 00:00:00 to the last day at 23:59:59. Presets: today, yesterday, last 7 days, this month, last month, last 30 days, last 90 days, this year. A period covers at most 370 days |
| Time zone (UTC) | The clock the start and end are read on — and on which hours, days and weeks are cut in the charts. It starts on your browser’s time zone; pick another to read the operation on a branch’s clock |
| Channel | All, WhatsApp or FlyB Chat |
| Number | With WhatsApp selected, one of your WhatsApp numbers |
Picking a preset, a time zone, a channel or a number updates the screen at once; typed dates wait for Apply. Long periods take a few seconds to compute — the screen says so, and keeps the previous figures on screen, dimmed, while the new ones are calculated. A figure computed less than a minute ago is reused rather than recalculated; Refresh shows when the figures were computed.
The address bar keeps the filters, so a link to the screen opens the same view for a colleague.
Now
The strip above the figures is the queue right now, whatever the period: how many customers are waiting for an answer, the longest wait, how many of them have nobody assigned, how many have waited more than 15 minutes and more than an hour, and how many conversations are waiting on the customer or snoozed. It follows the channel and number filters. It turns red while anybody has waited more than an hour.
How a message is classified
Every figure on the screen reads the same rules:
- A person’s message — written by somebody on the team, in FlyB or typed on the number’s own phone (WhatsApp Business). An automation’s message that a member chose to send from the Inbox counts as theirs. A message a Notes & Tasks task sent on a schedule does not.
- A bot reply — an answer the AI sent from your answer catalog, or from a message automation that answers incoming messages (first message or message received). Booking reminders, CRM stage messages and other automations are automations, but not replies.
- A failed send never counts as an answer: it never reached the customer.
The headline figures
Each figure shows its change against the previous period of the same length — the period just before it. While the period is still running (this month, on the 17th), only the part already elapsed is compared, against the same stretch of the previous period: half a month is never compared with a whole one. For times and rates, lower is better and the change is coloured accordingly; rates change in percentage points (p.p.).
| Figure | What it measures |
|---|---|
| First response (FRT) | The average time from the customer’s first message in a conversation to the first message written by a person. Bot replies do not count. Counted for the conversations whose first customer message falls in the period; the median and the 90th percentile are shown beside the average |
| Handling time (AHT) | The average time from the first message a person wrote in the conversation to its resolution. Counted for the conversations resolved in the period that somebody on the team wrote in — so it depends on the team resolving conversations |
| Conversations received | New conversations the customer started. Those the business started — a template, an automation, a team message — are counted apart, with the share that got a reply |
| Conversations resolved | Conversations resolved in the period, with how many were still pending at its end |
| Handoff rate | Of the conversations a bot answered before any person, the share where a person had to step in — because they wrote afterwards, or because the customer is still waiting. The rest were handled 100% by the bot |
| Abandonment rate | Of the conversations where the customer wrote and a person answered for the first time in the period, the share where the customer never wrote again in that conversation. A conversation no person wrote in is not counted: an automation may have solved it |
| Reopen rate | Of the conversations resolved in the period, the share where the customer wrote again on the same number (or the same chat) within 48 hours — a sign of conversations closed too early |
| Outreach with no reply | Of the WhatsApp templates sent in the period, the share that got no reply on the same number within 24 hours. Failed sends are left out |
Below them: the average reply time (every wait, not only the first — from each unanswered customer message to the next message from a person), concurrent conversations per attendant, the messages received with the business’s own messages split by who wrote them, the conversion rate, and the attributed revenue (with permission to view Financial).
Rates are only judged once they can be. A conversation answered ten minutes ago has not been abandoned yet: it joins the abandonment rate 24 hours after the answer, or as soon as it is resolved. A resolution joins the reopen rate after 48 hours (or as soon as the customer comes back), and a template joins the reply rate after 24 hours (or as soon as it is answered). The number still waiting to be judged is shown beside each rate. Without this, every report of “today” would look worse than the same day will look tomorrow.
Total service volume
Per hour, day, week or month (the grouping adapts to the length of the period, and can be changed):
- Columns — new conversations, split by who started them: the customer (received) or the business (started).
- A line — conversations resolved in each period, on the same scale.
- Below, on its own scale — the conversations pending at the end of each period: those the customer wrote in that were not resolved yet. It is the backlog: if it grows, the team is finishing fewer conversations than arrive.
Switch to Table to read every number.
When customers write
- Messages received by hour of day — 24 columns, one per hour, summed over the period. The tooltip gives the average per day.
- Conversations started by hour of day — the hour of the first message of each new conversation, split between the customer and the business.
- Heat map: weekday × hour — customer messages by day of the week and hour; the darker, the more messages. It is the picture to plan the team’s shifts from.
All three are read on the time zone chosen in the filters.
First response time, in bands
How long customers waited for the first person — up to 5 minutes, 5 to 15, 15 minutes to an hour, 1 to 4 hours, 4 to 24 hours, more than a day — with the abandonment rate of each band beside it. If abandonment climbs with the wait, customers are leaving because of the delay; if it is flat, they simply had nothing more to say.
Below the bands: how many conversations of the period still have no answer from a person, and how many have no answer at all, not even from a bot.
AI and automations
The handoff card splits the conversations a bot answered first into three: handled by the bot (nobody on the team wrote and nobody is waiting), handed over to a person, and waiting for a person right now. With the Answer Catalog module on, switch between AI + automations and AI only.
Performance by attendant
One row per person who wrote in the period’s conversations, sortable by any column:
| Column | |
|---|---|
| Conversations · Messages | The conversations they wrote in, and the messages they sent |
| FRT | The first response time of the conversations they answered first |
| Avg. reply | The average time of all their replies |
| AHT | The handling time of the resolved conversations they took on (the first person who wrote in them) |
| Resolved | The conversations they resolved |
| At once | How many conversations they kept going at the same time, on average, while working — and the peak. A conversation is active from their first message to 5 minutes after their last one, joining messages less than 30 minutes apart. The green band marks 3 to 5, the usual pace on WhatsApp |
| Now | Open conversations assigned to them right now, and how many of those are waiting for an answer |
| Abandoned | The share of customers who never wrote again after their first answer |
| Conversion | The contacts they talked to who were then won, paid or became qualified |
| Revenue | Payments from the contacts they talked to, made after their first message (with permission to view Financial) |
Replies typed on the number’s own phone have no FlyB user behind them and get a row of their own. Somebody who has since left the team keeps their history.
Credit is shared, not split: a contact two people talked to before it was won counts for both, so the rows can add up to more than the team’s figures, which count each contact once.
Active outreach
The WhatsApp templates sent through the official API in the period, as a funnel: sent, delivered, read, replied within 24 hours on the same number — with the failed sends apart and the most used templates, each with its read and reply rate. Delivered and read follow WhatsApp’s own receipts: a customer who turned read receipts off never shows as read. Numbers linked through WhatsApp Web do not send templates, and FlyB Chat has none.
Conversion and revenue
Each contact somebody on the team talked to in the period counts once. A contact is converted when, after the team’s first message and by the end of the period, it was won in the CRM, paid (a confirmed payment in Sales) or became qualified (MQL/SQL). Bookings made after the conversation are shown beside it, without entering the rate. Each part appears only with its module on.
Qualified contacts count from when FlyB started recording the moment a contact is qualified (October 2026): a contact qualified earlier has no recorded moment, so it cannot be credited to a conversation.
Attributed revenue adds up the confirmed payments of those contacts, made after the first message, per currency, with the revenue per conversation a person handled.
Channels and numbers
Each WhatsApp number and FlyB Chat with its conversations, messages received and messages sent — and the business’s messages split between people, automations and AI: how much of the talking the machines already do.
Who sees what
The screen counts over the conversations you could open in the Inbox:
- A WhatsApp number restricted to some members counts only for them, and a disabled number for nobody.
- A member who sees only their own contacts reads figures counted over their own contacts alone.
So two people can read different figures for the same period. Revenue needs permission to view Financial.
On the dashboard
The main figures are also dashboard widgets, in the Customer service group: first response time, handling time, conversations received and resolved, messages received, customers waiting now, and the handoff, abandonment, reopen and outreach-with-no-reply rates. They use the same definitions, on the dashboard’s own periods, and are offered to the members who can open this screen.