Guides

How to Categorize Support Tickets: 8 Categories, P1 to P4

Vera Sun

Last update

Summary

  • Sort every ticket by what the customer needs into eight categories, one of them for questions your docs never answered.

  • Rank each ticket P1 to P4 by how many customers are blocked and whether they have a workaround, and give every level a first-response target.

  • Let an AI agent answer the documented questions with the source on every reply, and keep the queue for the tickets that need a person.

To categorize support tickets, sort your last 100 by what each customer needed, into eight categories that each have an owner who acts next. One of the eight holds the questions your docs never answered.

Then score every ticket on two things: how many customers are blocked, and whether they can work around the problem. The two answers give four levels, P1 to P4, each with a first-response target.

The scheme earns its keep the week the queue doubles and half of it is marked urgent. An AI customer service agent that answers from your help center then takes the documented questions off the queue.

Why do ticket categories fall apart as the queue grows?

Ticket categories fall apart for three reasons, and each one shows up in the monthly report before anyone sees it in the queue.

  1. Product-area names: “Dashboard,” “Widget,” and “API” say where a problem sits, and nothing about who acts next. A billing question asked from the dashboard and a crash on the dashboard land in the same bucket.

  2. Urgency inside the type: Tettra’s 2024 list of the seven most common help desk ticket categories includes “Emergency tickets.” An outage on the billing page then fits Emergency and Billing at once, and two teammates file it two ways.

  3. No home for the docs gap: A question your help center never answered gets filed as a how-to. Someone answers it by hand, and the same question comes back next week.

The third reason is the biggest. In a random sample of 5,000 unresolved conversations from Wonderchat’s study of 5,848,078 conversations, 58.5% asked for an answer never documented anywhere the agent could reach.

Most unresolved conversations asked something no page had ever answered.

Out-of-date content explained only 1.2% of the sample, against 58.5% for answers nobody wrote. That gap is why the missing answer needs a category of its own.

A category is a promise about who acts next. If its rule can’t tell you that, rewrite it.

Which categories should a B2B support team use?

Use eight support ticket categories, each with an owner who acts next, such as support, billing, engineering, product, sales, or whoever owns the docs.

Category

Rule

Who acts

Example question

How-to

The docs already answer it

Support, or the AI agent

“How do I add a teammate to the inbox?”

Account and access

Logins, seats, users, or data moving between accounts

Support

“Can I move all my chat history to a colleague’s account?”

Billing and plans

Someone has to change an invoice, a payment, or a plan

The billing owner

“Can you reissue last month’s invoice with my new VAT number?”

Bug

Something broke; an outage is a Bug at P1

Engineering, through support

“The dashboard won’t load for anyone on our team.”

Integrations and API

A connected tool or the API fails

Support, then engineering

“Our API key returns a 401 since we rotated it.”

Feature request or feedback

The customer asks for something the product doesn’t do

Product

“Can we get a weekly CSV of tickets by category?”

Buying question

Plans, fit, or security before a purchase

Sales

“Do you sign a DPA before a paid plan starts?”

Not in the docs

One reply settles it, and no help center page says it

Whoever owns the docs

“Is there a status page we can subscribe to for outage alerts?”

When a ticket fits two categories, ask who has to act. If someone must change the account, it’s Billing and plans or Account and access.

If one reply settles it and no page says it, it’s Not in the docs. “Can I split one invoice across two payment cards?” goes there, while the VAT reissue goes to billing.

Ask the questions in order, and the first yes names the category.

Buying questions arrive in the support queue anyway, and they get their own category with sales as the owner. In the same study, support was 53% of all conversations and pre-sales 23%, across 5,000 scored conversations classified by stage.

Not in the docs is the category that improves your help center.

Luke Tan, CTO and co-founder of Wonderchat, decides by how often a question comes back. “If a question is asked many times… that usually indicates that a dedicated article is helpful,” he says.

An existing article with wrong or outdated information needs a fix instead. A one-off question gets a correction, in his words “a small one-off FAQ,” which “is usually good for extremely specific questions or edge cases.”

Keep one category per ticket. The product area goes in a tag, never in the category name.

How do you build categories from your own tickets?

Build the list from your own tickets in six steps and one afternoon.

  1. Pull your last 100 tickets

  2. Write what each customer needed in five words or fewer

  3. Sort each line into one of the eight categories

  4. Merge any category you used only once or twice

  5. Write a one-line rule and an example for each

  6. Check the list against next week’s tickets

Before you start: Have the last 100 tickets from your help desk or shared inbox, a sheet with four columns (the customer’s words, the need, the category, and the priority), one afternoon, and whoever owns the queue.

1. Pull your last 100 tickets

Take the most recent 100, across every channel that creates tickets. Older tickets describe a product you’ve since changed.

For each one, copy the date, the channel, the subject, and the customer’s first message. That message is where the need sits.

If your ticket volume is under 100 a month, take the last 60 days instead.

2. Write what each customer needed in five words or fewer

A subject line reading “Help!!” turns into “Invoice in Swiss francs.”

Write the need. The need decides who acts, and the subject line almost never says it.

Customer’s words (Wonderchat’s review inbox)

Need, five words or fewer

Category

Priority

“Can I pay my invoice by bank transfer in Swiss francs?”

Invoice paid in Swiss francs

Not in the docs

P4

“Can my invoices go to two email addresses at once?”

Invoices to two email addresses

Billing and plans

P4

“Is there a status page we can subscribe to for outage alerts?”

Status page for outage alerts

Not in the docs

P4

“Can you reissue last month’s invoice with my new VAT number?”

Invoice reissued with new VAT

Billing and plans

P3

3. Sort each line into one of the eight categories

Put every line in exactly one category. Whatever won’t fit shows you the category you’re missing.

4. Merge any category you used only once or twice

Out of 100 tickets, a category with two of them holds 2%, and thin categories turn the monthly report into noise.

Fold anything under 3 into the nearest category.

Look at the other end too. Split a category holding more than 30 of the 100 when two different people act on its tickets.

5. Write a one-line rule and an example for each

Write the rule before you share the list. Two people must file the same ticket the same way.

Hand ten tickets to a teammate, and if more than one lands somewhere else, rewrite that rule.

6. Check the list against next week’s tickets

Run next week’s tickets through the list before you call it done. Fixing it now costs an afternoon, and fixing it after a quarter costs every report since.

Watch two counts. Tickets that fit no category should be zero, and the share under Not in the docs is how much of the queue a help center page could answer.

That share answers the question support leads ask on Wonderchat’s sales calls, how much of their queue is the same few questions.

How do you prioritize support tickets?

Prioritize every ticket on two questions: who is blocked, and is there a workaround. That’s the impact and urgency check at the heart of ticket triage, and ticket prioritization fits in four levels.

1. Count the customers who are blocked

Count first. One blocked customer and fifty need different first moves.

A fourth ticket in an hour about the same broken dashboard counts as many, even when each ticket reads like a single complaint.

2. Check whether they have a workaround

Ask before you reply. The level should follow the customer’s day.

A customer who can keep working with a workaround can wait a day, and a customer who can’t do their job can’t.

3. Match the two answers to a level

Put the two answers side by side, and one of four support ticket priority levels follows.

Level

Who is blocked

Workaround

Example

First-response target to start

P1

Many customers, or data at risk

None

“The dashboard won’t load,” the fourth ticket this hour

1 hour, then updates until it’s fixed

P2

One customer

None

“Our API key returns a 401 since we rotated it.”

4 business hours

P3

One customer

Yes, or they can keep working

“Can you reissue last month’s invoice with my new VAT number?”

1 business day

P4

Nobody

Not needed

“Is there a status page we can subscribe to for outage alerts?”

2 business days

Wonderchat’s helpdesk sorts tickets into three priorities, and the four levels map onto them. In Wonderchat, open Inbox, select the settings gear, then the Priority tab in Ticket settings.

High covers a customer who “is blocked or at risk,” so it holds P1 and P2. Medium holds P3, and Low holds P4, which the app describes as “General questions, how-to requests, feature requests, feedback, and anything informational with no urgency.”

Each definition can be edited to match your levels. Turn on Set priority automatically for new tickets, and a priority someone set by hand stays as it is.

4. Set a first-response target for each level

Start with OneDesk‘s targets of 1 hour, 4 business hours, 1 business day, and 2 business days, which it calls “a common starting point for business-hours support teams.” Keep them for one month, then set each level’s target from your own median first response time.

Use the median. One slow ticket drags the average and barely moves the median.

Take a week of seven first replies, where six took 4, 6, 8, 10, 12, and 15 minutes and one took 9 hours.

The average is 85 minutes, and the median is 10.

A target set from the average would promise 85 minutes, while most replies took about 10.

Set first-response targets now, and resolution targets after a month of data. A resolution target you guess today becomes a promise you break next week.

The customer’s plan breaks ties inside a level, and it never lifts a ticket a level. Jitbit argues that customers who pay more “may deserve a higher priority tier,” and OneDesk writes “Customer tier adjusts targets, not priority.”

Side with OneDesk. Once a plan can lift a level, an enterprise customer’s how-to question jumps ahead of a free customer’s outage, and P1 stops meaning blocked.

What does a sorted queue look like, ticket by ticket?

Sorted, the ten tickets give one P1, one P2, two P3s, and six P4s.

Customer’s question

Category

Priority

Why

First move

“The dashboard won’t load,” the fourth ticket this hour

Bug

P1

Many customers blocked, no workaround

Open an incident and send one status update to everyone affected

“Our API key returns a 401 since we rotated it.”

Integrations and API

P2

One customer blocked, no workaround

Check the key and the logs, then hand it to engineering

“Can I move all my chat history to a colleague’s account?”

Account and access

P3

The docs answer part of it

Send the documented part and ask what the rest is for

“Can you reissue last month’s invoice with my new VAT number?”

Billing and plans

P3

The product works, the invoice is wrong

The billing owner reissues it

“Can I get an invoice addressed to two different companies?”

Billing and plans

P4

A request, nobody blocked

The billing owner answers, then adds the answer to the billing page

“Can I split one invoice across two payment cards?”

Not in the docs

P4

One reply settles it, no page says it

Answer it, then write it into the billing page

“Is there a status page we can subscribe to for outage alerts?”

Not in the docs

P4

Nobody blocked, no page answers it

Answer it, then write the page

“Can we get a weekly CSV of tickets by category?”

Feature request or feedback

P4

Nobody blocked

Log it for product with the customer’s reason

“Do you sign a DPA before a paid plan starts?”

Buying question

P4

A buyer waiting, nobody blocked

Hand it to sales the same day

“How do I add a teammate to the inbox?”

How-to

P4

The docs answer it

Send the help center page

A sample queue: five questions come from Wonderchat’s own review inbox and product screens, and five are written for the categories those don’t cover.

Three of the five real questions are about invoices and payments. None of them needs engineering, and each one needs a billing answer written down once.

Once a ticket has its category and level, how to answer support tickets covers the reply itself.

What changes when an AI agent answers first?

An AI agent changes which tickets reach the queue at all. The agent answers the documented how-to, account, and billing questions from your help center and docs, and links the page each answer came from.

When no page covers a question, it says it can’t find that, doesn’t guess, and offers a person.

The ticket arrives with the full chat attached. You sort from the customer’s own words in every message they sent.

Not in the docs then fills itself. In Wonderchat, open Review, then the Inbox tab, then Knowledge gaps, and each gap shows how many times it was asked, with its source chats.

Every question the agent couldn’t answer waits in one inbox until someone gives it a source.

Answer each gap with Paste a URL, Pick from Files, or Write it myself, and the agent answers that question from then on. In Wonderchat’s study, 84 businesses that added missing documentation went from 75.4% to 80.8% resolved in three months, and 57% improved while 23% got worse.

Wonderchat’s support team sees one failure more often than a made-up answer. The agent finds an archived policy and answers from it correctly, word for word, with a rule that no longer applies.

File those under Not in the docs too. The fix is retiring the old page, so the agent stops reading it.

Count Not in the docs every Friday. That count falls as the docs get better, which makes it the number to show your founder.

Should you tag tickets by hand, with rules or with AI?

Tag by hand for the first month, then add rules, then AI. Each stage needs the one before it to hold.

Stage

When to start

What it costs

What breaks if you start too early

By hand

Day one, for the first month

Someone’s time on every ticket

Nothing, and it’s how you find the missing category

Keyword rules

Once a category’s words repeat, such as “invoice,” “VAT,” or “401”

An afternoon to write the rules

“Invoice not received” and “invoice in Swiss francs” land together

AI tagging

Once the list stops changing

The plan that includes it

It learns a scheme you’ll change next month

In Wonderchat, a teammate can add a tag to any ticket by hand, from Tags and then Add in the ticket’s details. A conversation can be assigned to a teammate by hand too.

The Scale plan adds auto-tagging and rules for routing and assignment. Each tag carries its own rule, written in plain words, and the agent applies it to matching chats.

A tag’s rule reads like a category rule, and on Scale the agent applies it to every matching chat.

Start with what to automate first before you pay for tagging. A list that still changes every week costs more to automate than to tag by hand.

When does your current help desk already do this?

If your help desk ticket management already runs on Zendesk triggers, keep Zendesk. Its pricing page lists “Automations and triggers” and “Ticket routing” on the Support Team plan, so rules-based sorting may already be in the help desk you pay for.

Wonderchat works on top of Zendesk with ticket sync from Basic, up to 50 tickets, and no cap from Scale, and nothing moves. The same goes for Freshdesk, with its full help desk and omnichannel suite.

Teams can use Wonderchat’s helpdesk from the first day or keep their own. Pick Wonderchat’s when you’re building your first queue, and keep Zendesk or Freshdesk when their rules already sort your tickets.

How does Wonderchat handle the tickets in your queue?

Start with the scheme, eight categories that each have an owner and four levels by who’s blocked. Then let Wonderchat, an AI customer service agent and AI helpdesk in one, take the documented questions.

Take the status-page question. The agent says no page covers it and offers a person, the ticket lands in the shared inbox with the full chat, and the question waits under Knowledge gaps until someone writes the page.

Tickets, the shared inbox, and live chat come on every plan, Free included. Routing and assignment rules and auto-tagging come with Scale (see plans).

Start for free, paste your help center URL, ask your hardest question, then one your docs can’t answer, and see the source on every answer and the honest miss, or book a demo.

Frequently asked questions

What are the categories of tickets?

Use eight, each owned by the person who acts next. They are How-to, Account and access, Billing and plans, Bug, Integrations and API, Feature request or feedback, Buying question, and Not in the docs.

How do you prioritize tickets?

Count the customers who are blocked, then check for a workaround. Many blocked with no workaround is P1, one blocked is P2, one with a workaround is P3, and nobody blocked is P4.

What are the different categories of tickets in ITIL?

ITIL sorts IT work into incidents, service requests, problems, and changes. Those types suit an internal IT desk. A customer queue sorts better by the customer’s need, and an incident there is a Bug at P1.

How many ticket categories should a support team have?

Six to eight, plus Not in the docs. Jitbit suggests five to eight and HelpSpot under 20. Past ten, two people start filing the same ticket two ways.

What are the best ticketing systems?

Look for tickets that carry the full chat, categories you can rename, and assignment to a teammate. Start with support ticketing systems compared, which ranks nine.

AI answers from your docs. Tickets for the rest. Free to start.

Vera Sun

Vera Sun is the co-founder of Wonderchat, an all-in-one AI Support and Conversion agent platform built for companies with large knowledge bases. Her background is in product design and product management, which shapes how she thinks about the messy space between what a customer asks and what a product can answer. She writes about AI agents in production, customer support, and go-to-market for technical buyers, and can usually be found tinkering with new AI tools.