Quality
One question behind every score: did your team understand the customer's real problem and move it to a fix, with the least effort for the customer?

Each chat we score, on any channel, gets a score out of 100, and every point links to the message behind it. Eight things decide it: understanding, the right fix, ownership, accuracy, the customer's effort, tone, speed, and checking it's really closed. A greeting or a “sorry” earns nothing on its own.
Mood: worried → resigned
Red lineOnce you approve the fix below and share it with your team, the same dodge is a red line: 0/100, with every finding kept.
Criteria (earned / weight)
Solution Bank
Looking up: “my order hasn't arrived”
Suggested, waiting for your OKApproved and shared with your team
Learned from a chat Ella solved (score 96, not reopened in 7 days)
Doesn't work: “wait” without checking.
Promise limit: no compensation without your OK.
Better reply
“I've opened order 4471: it's been with the courier since Thursday, and here's your tracking number. I've sent it to our shipping team with the order number, and I'll message you a delivery date before 5 pm. If you'd rather cancel, you'll get your money back under our policy.”
Coaching tip
Late order? Open it by number, give tracking and a date, and transfer it with its number. Never just “wait”.
After this chat
So every score comes with three lights: Customer (was it solved?), Business (order number, ticket, next step) and Compliance (red lines). In one published case, an overall quality score of 86% hid accuracy of just 60% on what mattered to customers.7
“Restart your phone” counts only if it fits the problem, is followed by “did it work?”, and ends with a ticket number and a time. Otherwise it's a dodge, and it costs points even before your Solution Bank has the fix. So do “please wait” with no check, and blaming the courier with no follow-up.
Not every problem is solved inside the chat. Sending the request to the right department is a real fix, when the customer gets the order or ticket number and knows when they'll hear back. That's ownership. A transfer with no number, or “call our other number” with no follow-up, loses points: the customer will have to explain it all again.
Please wait a bit, the delivery company is slow these days.
I've sent it to our shipping team with order number 4471, and I'll get back to you with a date before 5.
Most dodges aren't laziness. A target that rewards closing chats fast makes “restart your phone” tempting. An agent who doesn't know the fix guesses. An agent with no access to order tracking can only say “wait”. Airest labels each dodge (training, tools and access, or policy) so you fix the cause, not just the person. And soon the right fix from your Solution Bank appears to your agent while the chat is still open.
Repeat contacts: the same customer, about the same problem, back within 7, 15 or 30 days on any channel. A chat whose customer comes back within 7 days is marked failed after the fact, so a polite script can't fake a good score. Under 10% within 7 days is good, under 5% is excellent, and over 15% calls for questions.10 In call centres, every 1% gained in first-contact resolution brings 1% more customer satisfaction.2
Customers back within 7 days, about the same problem
When a problem comes up 3 times or more, Airest picks the chat that solved it best (a score of 90 or more, not reopened within 7 days, no bad rating), writes the fix into your Solution Bank and waits for your OK. Write fixes yourself any time; day one starts with your industry's common problems. The red line for skipping a known fix applies only to fixes you approved and shared with your team.
We read the customer's mood at the start and at the end, and judge the change, not their first angry message. Warmth at the start and end with efficiency in the middle lifts satisfaction and spending,12 and specific words (an action, an item, a date) beat general ones.13 We never require “sorry” or a greeting: “thank you for waiting” often works better than “sorry”, but a serious mistake needs a real fix first.14 Warm dialect is not unprofessional.
Each industry pack changes the weights, the speed target and the red lines. Where the law sets the bar, the pack follows it, and every new pack passes a legal review before it opens.
Order number and status, then tracking or a new date. Returns exactly as your written policy. “In stock” only after checking. A wrong price is a red line.
No diagnosis and no dose in chat; emergencies are sent on at once.
Allergy answers come from your written allergen list, never from memory, and “I can't confirm that” is a good answer.15
The VIN, or year, make, model and engine, before “it fits your car”; then the exact part number, and whether it's genuine or aftermarket.16
“Restart” only when it fits the symptom, then “did it work?”, then a ticket with a number and a time. Collect the details your Solution Bank asks for before transferring, so the customer never explains twice. Never ask for a password. A new fix is written in the customer's own words, so the next agent finds it.11
Identity is checked before any result,34 and results are never interpreted in chat: that's for the doctor. Ready times are realistic. Health data is sensitive personal data under Qatar's and Saudi Arabia's data protection laws.3536
Never ask for a PIN or a one-time code. Taking a full card number in chat puts the whole channel under the card industry's security standard,18 so card details go to the official channel, and sending a customer there with the reason is not a dodge. Complaints get a reference number and a time frame. Qatar's central bank requires licensed firms to tell customers when they're dealing with AI.37
Insults, sharing another customer's data, a promise outside your policy, leaving a customer with no reply, and dodging a fix you approved: each is a red line, and each industry adds its own (a wrong price, a diagnosis, a guess about allergies). One red line makes the score zero. Every finding stays visible, and you can raise the score with a reason.
One red line: 0/100
Every point is tied to its message. A customer's angry first message never counts against the agent. Nobody is ranked before 15 chats in a week, and only you see the ranking. Raise any score with a reason. Flag what it missed, and it learns your standard. Tell your team before you start: this is coaching, not policing.
An optional “How was our service today?” after the chat, from your own number, about the agent who served them. You switch it on; agents never see the answers. The survey and our score measure different things: a study of more than 265,000 tickets found no link between them,20 so you see both side by side. When they disagree, a high score with an unhappy customer goes to the top of your weekly report: that's where our scoring, or your Solution Bank, may be wrong.
Example. Names and numbers are illustrative. Nobody is ranked before 15 chats in a week, and only you see the ranking.