Spot a new problem in the first 20 tickets, not 2,000
Tags each new ticket with symptom, service area, platform and whether it followed a change. Your code counts the tags by the hour and raises the alert.
Try it on this example
Ticket subject: Can't pay for my order
Ticket text, with any replies from the customer
- What is going wrong for the customer?Payment fails100%
- Which part of the service is the ticket mainly about?Payment methods100%
- On which platform does the customer say the problem happens?iPhone app100%
- Does the customer say the problem began recently?Yes99%
- Does the customer link the problem to a change?After an app update99%
- How seriously does the problem affect the customer?Blocked100%
- Does the ticket say that more than one person has the problem?No93%
- Does the customer say what they have already tried to fix the problem?Yes98%
- Does the customer quote an error message or error code they saw?Yes93%
- Is there enough in the ticket to tell what is going wrong?Yes95%
These are real answers stored from one run on this example.
The prism behind it
Spot a new problem in the first 20 tickets, not 2,000
Fields
- Ticket subject
- Ticket text, with any replies from the customer
Context
We are Basketwise, a grocery delivery service with an iPhone app, an Android app and a website. Customers contact us by in-app chat, email and the help form. Every new ticket is read here when it is created. Code counts tickets by symptom, part of the service and platform every hour, compares the count with the usual rate, and alerts the incident channel on a spike. Code also matches the tags against the open incidents and sends the status reply for a known incident. These answers only tag the ticket; they never count, match incidents or name a cause. Parts of the service: - Sign-in and account: signing in, passwords, codes, account details and addresses. - Browsing and search: finding products, prices shown, product pages. - Basket and checkout: the basket, choosing a delivery slot and placing the order, when paying is not the problem. - Payment methods: adding, choosing or using a card, a digital wallet or a gift card to pay. - Delivery and tracking: the driver, delivery times, tracking and missing or damaged bags. - Items and substitutions: wrong, missing or substituted items, and product quality. - Offers and rewards: vouchers, promotions, the loyalty scheme and delivery passes. - Notifications: push notifications, emails and texts from us.
Questions
What is going wrong for the customer? Choice
Pick what the customer sees go wrong, not what they think the cause is. When the ticket raises several problems, pick the one that stops them most.
Which part of the service is the ticket mainly about? Choice
Use the parts of the service in the context. Pick the part where the problem happens.
On which platform does the customer say the problem happens? Choice
Go by what the customer says, such as "on my iPhone", "the app on my Samsung" or "on your website". Do not guess from the wording.
Does the customer say the problem began recently? Yes / No
Count relative words such as "today", "this morning", "since yesterday", "since the update" or "it worked last week". A date on its own does not count, because code compares dates. Yes: The customer says, in relative words, that the problem is new. No: The customer says it has gone on a long time, or says nothing about when it began.
Does the customer link the problem to a change? Choice
Go only by what the customer says, not by what might have caused it.
How seriously does the problem affect the customer? Scale
Judge from what the customer says they cannot do, not from how upset they sound.
Does the ticket say that more than one person has the problem? Yes / No
Count a household member, a friend, a neighbour, "everyone" or posts from other people the customer has seen. Yes: The ticket says or clearly implies that more than one person has the problem. No: Only the writer is affected, or the ticket says nothing about anyone else.
Does the customer say what they have already tried to fix the problem? Yes / No
Count restarting the app or phone, reinstalling, signing out and back in, trying another card or another device, or any other step the customer says they took. Trying the same thing again does not count. Yes: The customer says they tried at least one step. No: The customer mentions no step they took.
Does the customer quote an error message or error code they saw? Yes / No
Count the words of an on-screen message or a code, quoted or copied. "It says error" with no words from the message does not count. Yes: The ticket quotes an error message or code. No: No message or code is quoted.
Is there enough in the ticket to tell what is going wrong? Yes / No
Answer No when the ticket is only a subject line, a few words such as "not working", a screenshot with no text, or too garbled to read. Yes: A careful reader could tell what is going wrong for the customer. No: The ticket does not say enough to tell.
Lens columns
symptom, symptom_probability, product_area, product_area_probability, platform, platform_probability, recent_start, recent_start_probability, after_change, after_change_probability, severity, severity_average, many_users_affected, many_users_affected_probability, already_tried_fix, already_tried_fix_probability, error_message_quoted, error_message_quoted_probability, enough_information, enough_information_probability
Run it on your own text
Add this prism in the app, change any question, and test it on a file of your own.