Skip to content
Chatbotscape
Editorial flat-vector illustration for How to Set Up Chat Tags for Reporting
Published 8 min read

How to Set Up Chat Tags for Reporting

Quick answer: Start from the questions you want answered, not from a list of tags. Read a sample of real chats, group them into a short list of contact reasons with one-line definitions, and make the main reason a required pick-list field where your platform has one. Let rules add the tags the system can already know, such as channel or bot handoff. Remind agents to tag before closing, then read one report a week and prune tags nobody uses.

Step 1: Decide what the tags must answer

Write down the three or four questions you want the data to answer every week. Typical ones for a small team:

QuestionWhat you need to tagWho reads it
Why do customers contact us?One main contact reason per chatOwner, support lead
What could the bot or help center handle?Reason plus whether the bot handed overWhoever maintains the bot
What is breaking right now?Short-lived flags such as an outage or a bugProduct or operations
Which chats need a follow-up?A flag cleared when the follow-up is doneAgents

If a tag does not serve one of these questions, you do not need it yet.

Step 2: Build a short list from real chats

Do not invent the list in a meeting. Open your archive or chat transcripts and read 50 to 100 recent conversations. For each one, write the reason in plain words, then group the notes. You will usually find a handful of reasons covering most chats and a long tail of one-offs.

Rules for the list:

  1. One idea per tag, written as a definition. "order_status: the customer asks where an order is or when it arrives."
  2. Consistent format. Zendesk only accepts letters, numbers, dashes, underscores, colons and forward slashes, and treats late_order and late-order as different tags, so pick one connector and stick to it. LiveChat allows tag names up to 70 characters.
  3. An other option, so agents never force a chat into the wrong bucket. If other grows, read those chats and add a reason.
  4. Keep it short. No source we read gives an ideal count; the test is whether an agent can pick in a few seconds. For the separate job of reviewing a weekly sample to improve a bot, our conversational analytics guide works with just five tags.

Step 3: Use a required field for the main reason, tags for the rest

Free-form tags drift. Where your platform has structured fields, put the main contact reason there and keep tags for extra flags:

  • Intercom separates conversation attributes (structured, such as a predefined list, and usable as required fields before a conversation closes), topics (automatic, keyword-based, built for trend reports and applicable to past conversations) and tags (flexible, applied to individual replies). Intercom recommends combining all three. See the Intercom review.
  • Zendesk uses tags across tickets, users and organizations, and tags on a user or organization are copied onto their tickets, which is handy for flags such as a key account. See Zendesk.
  • LiveChat lets admins limit a tag to one group so only that group's agents see it, which keeps each team's list short. See LiveChat.

Step 4: Automate the tags the system already knows

Anything the platform can know without judgement should not depend on an agent remembering:

  • Rules. Intercom Workflows can add a tag when a message contains words like "bug" or "error", or when a conversation arrives from a channel such as WhatsApp. Zendesk triggers, macros and automations can work with tags too.
  • Keywords. Zendesk's automatic ticket tagging adds tags from words in the ticket description. Its own warning: common words such as "and" or "the" make it much less useful, so check what it produces.
  • Suggestions. LiveChat starts suggesting tags once at least 100 chats are tagged, on its Team, Business and Enterprise plans.
  • Bot handoff. If your bot can add a tag when it passes a chat to a person, add something like bot_handoff. It shows which reasons the bot cannot finish, which is the core of improving resolution rate.

Before trusting any automatic tag, compare it with what a person would pick on 20 or 30 chats. A wrong tag applied to every chat is worse than no tag.

Step 5: Keep agents consistent

Most tagging fails here, on busy days:

  • Tag before closing, not from memory at the end of the shift. LiveChat has an optional reminder that asks agents to tag a chat when they archive it; Intercom attributes can be made required before closing.
  • Pin the common tags. LiveChat lets agents pin frequent tags to the top of the list; Zendesk's autocomplete suggests the 15 tags most used on tickets in the last 60 days.
  • Share the definitions where agents work, and walk through five tricky chats together once a month.
  • Name one owner who adds, merges and retires tags. Without an owner the list only grows.

Tags can also drive work, not just reports: a billing tag added by a rule can feed routing chats to the right agent.

Step 6: Read one report a week and prune

Pick one view and read it on the same day each week:

  • LiveChat has a Tags usage report showing the most common topics, and other reports and Archives can be filtered by tag, with Match all or Match any for several tags.
  • Zendesk lets you build views and reports from tags and search tickets by tag.
  • Intercom includes topics, attributes and tags in CSV exports; it positions topics and attributes for trend reporting and tags for individual flags.

Look at three things: the top reasons and how they moved, the share of chats with no reason at all, and the reasons most often handed over by the bot. Then act on one of them, for example by writing a help article or a bot answer for the top reason. Once a month, retire tags with almost no use and merge near-duplicates. For deeper reading of the conversations behind a number, use the conversational analytics guide, and see the chatbot metrics guide for which numbers to put next to tags.

Check the arithmetic

An invented example, not a measurement. A shop handles 600 chats a month. In the first month agents tag 450 of them: 450 ÷ 600 = 75%. After turning on the reminder to tag before archiving, 570 of 600 are tagged: 570 ÷ 600 = 95%. The top reason, order_status, appears on 171 of the 570 tagged chats: 171 ÷ 570 = 30%. The bot handed 40 of those 171 chats to a person, about 23%, mostly when the customer had no order number. The shop adds an order-lookup step to the bot and watches the same three numbers the next month. These numbers are made up to show the method.

Common mistakes

  • Starting with 60 tags before reading a single chat.
  • Letting spellings multiply, so one reason shows up as three report lines.
  • Using tags for the main reason when a required field would give cleaner data.
  • Trusting auto-tagging blind, without checking it against real chats.
  • Never deleting a tag, so the list becomes unusable.
  • Collecting tags nobody reads, which costs agent time and changes nothing.

FAQ

How do I set up tags for customer support chats? Read a sample of recent chats, group them into a short list of reasons with written definitions, create them in your platform's tag settings, and decide which are added by agents and which by rules. Then review a report weekly.

Should agents tag during the chat or after it? Before closing it, while the conversation is fresh. Tagging from memory at the end of a shift produces guesses. A reminder on archive or a required field before closing helps.

How many tags should a small support team use? No source we read sets a number. Keep the list short enough that agents choose in seconds, include an "other" option, and retire tags nobody uses.

Can I tag chatbot conversations automatically by topic or intent? Yes, on many platforms: rules on message text or channel, keyword auto-tagging, or AI suggestions. Check automatic tags against what a person would choose before trusting the reports.

How do I find the most common customer questions from tags? Use your platform's tag report or filter, look at the top reasons each week, and read a few chats behind the biggest one before acting on it.

What is the difference between tags and conversation attributes? Tags are free-form labels you can add several of; attributes or fields offer a fixed set of values. Use a field for the main contact reason and tags for extra flags.

Sources

  • LiveChat, Tagging chats and tickets — livechat.com/help/tagging-chats-and-tickets/, read 14 November 2026. Source of tag creation, the 70-character limit, group tags, pinning, the archive reminder, suggestions after 100 tagged chats, the Tags usage report and tag filters.
  • Zendesk, Using tags — support.zendesk.com/hc/en-us/articles/4408888664474, read 14 November 2026. Source of tags on users and organizations, automatic ticket tagging, views, reports and workflow use.
  • Zendesk, Working with ticket tags — support.zendesk.com/hc/en-us/articles/4408835059482, read 14 November 2026. Source of allowed characters, multi-word tags, autocomplete from the 15 most-used tags and the common-words warning.
  • Intercom, How and when to use conversation topics, attributes and tags — intercom.com/help/en/articles/7126365, read 14 November 2026. Source of the three-way split, required attributes and CSV exports.
  • Intercom, Tag conversations automatically with rules — intercom.com/help/en/articles/6512873, read 14 November 2026. Source of Workflow tagging by message content and channel.
  • Chatbotscape evaluation methodology. /methodology (continuously updated).