Skip to content
Product
PlatformPricingDimensionsIntegrationsSecurityCompare
Solutions · by use case
Customer, guest and fan experienceProduct feedback and prioritizationRevenue, churn and retentionAgenciesContact center performanceAI agent and chatbot analyticsMulti-location performanceReviews and competitive benchmarking
Solutions · by industry
Hotels and hospitalityRestaurants and food serviceFitness and leisureRetail and e-commerceSports and live entertainmentInsuranceBanking and financial servicesSoftware and SaaSAutomotiveConsumer products and CPGContact centers and BPOsPublic sector and nonprofitReal estateGaming and betting
Learn
Artifacts GalleryCustomer storiesResearchBlogThought leadershipLessons Log in Start now
Blog

How to Find Top Contact Reasons in Gladly Across Voice, Chat and Email

Gladly's Topics report counts the Topics agents apply. Here is how to read the conversations themselves across voice, chat and email, and see where the two disagree.

To find the top contact reasons in Gladly across voice, chat and email, start with Gladly's Topics report, then check it against the conversations themselves: export the conversation items, put each conversation's customer messages and call summary into one text, label every conversation with its main reason from one fixed list, and rank the reasons by count and by what they cost, such as repeat contacts. Our Gladly integration page covers the connection itself; this page is about the analysis. For a support team that wants that read every week on every conversation in every channel, joined to order data and turned into reports for the teams that can fix the causes, Dimension Labs is built for the job.

What does Gladly's Topics report already tell you?

Quite a lot, if Topics are applied well. Gladly describes the Topics report as the place to "see the top reasons why your Customers were contacting you", counting created conversations tagged with each Topic. Three details shape how to read it:

  • Parents collect their sub-topics. A sub-topic is also counted in every higher-level Topic, so the broad parents usually top the chart. The CSV download gives both the roll-up and the directly applied counts; use the directly applied column when you compare reasons at the same level.
  • Dates follow the conversation, not the Topic. The report's date is when the conversation was created, not when the Topic was applied.
  • A conversation can close with no Topic. That is the default; an administrator can make a Topic required before closing. For conversations Gladly's AI handles, Automatic Topics can classify intent and apply a Topic, and Gladly says it labels conversations that are missing one.

Gladly's own maintenance advice points at the check this page describes: look regularly at what lands in "Other" and ask whether agents are applying the right Topics. Reading the text answers that question for every conversation at once, including the ones that closed without a Topic.

Where is the text for each channel in a Gladly export?

Gladly keeps every interaction with a customer on one Conversation Timeline, and a single conversation can span several channels, which makes it a good source for a cross-channel count. The export is where the words are. Gladly's help center points to its API for conversation transcripts: your organization gets an export file every 24 hours by default, each kept for 14 days. Each export job is a zip with four files: agents.jsonl, conversation_items.jsonl, customers.jsonl and topics.jsonl, the last one holding your Topic records and hierarchy. For history older than the files on hand, Gladly says multi-day historical raw transcript data comes from Gladly Support.

From Gladly's help center, read 2026-11-17

Chat and email text is in the export; a phone call arrives as a recording link and, with an add-on, a summary

What a conversation item in Gladly's export carries for each channel, and what to do with it before you count reasons.

ChannelWhere the words are, and what to do
Chat, SMS, social messaging
CHAT_MESSAGE
content.content on each message. Keep the customer's messages (initiator.type CUSTOMER) and join them into one text per conversation.
Email
EMAIL
content.content, with content.subject. Strip quoted replies and signatures so old threads are not counted twice.
Voice
PHONE_CALL
A recording link, and a text summary in recordingSummary when Voice Summaries is on. Use the summary, or transcribe the recording, and add it to the conversation as one more message.
Notes, Topic changes
NOTE, TOPIC_CHANGE
Note text, and the Topic IDs added or removed. Leave notes out of the reason text; keep the Topic IDs to compare against later.
Source: Gladly, API Export Tool, Help Customers on Different Channels and Review Text Summary of Phone Calls, read 2026-11-17.

Voice needs one decision. Gladly's Voice Summaries add-on writes a text summary of each recorded call about a minute after it ends and generates summaries in English. A summary is usually enough to name the reason for the call. If you need the customer's own words, for wording or effort, transcribe the recordings and send each call in as one more message in its conversation. We cover that step in turning call transcripts into structured data.

How do you rank contact reasons from Gladly conversation text?

  1. Make one row per conversation. Group conversation_items by conversationId, keep the customer's messages (initiator.type CUSTOMER) and the call summary or transcript, and drop notes and status changes. Gladly's own guidance is that a conversation can be about more than one thing, so keep room for a second reason. Remove card numbers, emails and phone numbers first; see removing PII from chat transcripts.

  2. Write the reason list from your Topic hierarchy. Start from topics.jsonl so the result maps back to what agents already use. Gladly suggests 2 to 3 levels and 5 to 15 sub-topics under a parent, with no overlapping Topics; the same rules make a good list for a model. Our post on building a contact reason taxonomy goes further.

  3. Label every conversation, then check a sample by hand. Ask for one main reason and at most one second reason from the list, with "none of these" allowed. Read a sample from each channel before trusting the counts; checking AI tagging accuracy covers how many to read.

  4. Count conversations, and compare with Topics. Rank main reasons by conversations, then put the agent-applied Topics next to them, channel by channel. The places they disagree are where Topic training or the hierarchy needs work.

  5. Rank by cost as well as count. Join each reason to what it costs: repeat contacts from customerId and dates, handle time and first contact resolution from Gladly's Conversation Export, and orders through customers.jsonl, which maps each customer to your own ID. Our post on which contact reasons take longest to resolve shows that ranking.

What changes when you check Topics against the text?

To show what step 4 looks like, we built a sample: 2,400 fictional conversations for an online retailer, 1,041 chats, 737 emails and 622 calls, each with agent-applied Topics and a main reason read from the text. The numbers are invented; the pattern is the one to look for in your own export.

In this synthetic sample, 362 conversations (15.1%) closed with no Topic, and calls made up most of the gap: 30.4% of calls had none, against 7.8% of chats and 12.5% of emails. Calls also carried more of the damaged and wrong item problems, so that reason lost the most: its Topic count was 278 against 417 conversations whose text named it, a third fewer. "Where is my order", the first Topic in the list, picked up stray labels and came out 26.5% higher by Topic than by text.

Worked example, synthetic sample data

Read from the text, damaged and wrong items move up a place; by Topic they lose a third of their conversations

Top six contact reasons in 2,400 sample Gladly conversations, ranked two ways. Left: conversations carrying each agent-applied Topic. Right: conversations whose text gives it as the main reason.

By agent-applied TopicBy reason read from the textWhere is my order663 conversationsReturn or exchange295 conversationsRefund status282 conversationsDamaged or wrong item278 conversationsChange or cancel order182 conversationsProduct question169 conversationsWhere is my order524Return or exchange423Damaged or wrong item417Refund status316Change or cancel order255Product question222Byagent-appliedTopicBy reason readfrom the textWhere is myorder663conversationsReturn orexchange295conversationsRefund status282conversationsDamaged orwrong item278conversationsChange orcancel order182conversationsProductquestion169conversationsWhere is myorder524Return orexchange423Damaged orwrong item417Refund status316Change orcancel order255Productquestion222
Sample data built for this post: a fictional online retailer, 2,400 conversations across chat, email and voice. Not any real company's figures.

Second reasons are the other blind spot. In the sample, 487 conversations raised a second issue, usually a refund question after a return or a damaged item, and agents tagged that second issue on 125 of them (25.7%). Once each sample reason is joined to what happened next, refund status stands out: in the sample, 31.0% of those conversations were followed by another contact from the same customer within seven days, against 16.0% across all reasons.

Worked example, synthetic sample data

Refund status is the fourth most common reason and the one most likely to bring the customer back

Share of conversations in which the same customer got in touch again within seven days, by the main reason read from the text.

Refund status316 conversations
31.0%
Damaged or wrong item417 conversations
23.3%
Return or exchange423 conversations
16.8%
Change or cancel order255 conversations
14.1%
Account or login105 conversations
10.5%
Where is my order524 conversations
9.4%
Promo or price138 conversations
8.0%
Product question222 conversations
5.4%
0102030
Same sample of 2,400 fictional conversations; all reasons together: 16.0% came back within seven days. In real data, take repeat contacts from customerId and the conversation dates in the export.

In a real export, the same three checks apply: how many conversations carry no Topic by channel, where the text and the Topics rank reasons differently, and which reasons bring customers back. The first two tell you whether to trust the Topics report as it is; the third tells you what to fix first.

Why would a Gladly support team read every conversation with Dimension Labs?

Dimension Labs is an omni-channel analytics agent for customer data. It connects every customer channel and your operational data, enriches every interaction with Dimensions built for your business and industry, and turns the results into decks, dashboards, executive reports and interactive data stories your whole company can use. Plans start at $299 a month with no demo.

For a Gladly team, that means chats, emails and SMS, plus the call summaries or transcripts you send in as messages, are read together with the reviews, surveys and other tools your customers use, so a damaged-item problem counts once whether it came in by phone or by chat. The Gladly connection sends conversation items through the universal API from Gladly's export files, and the integration side joins them to orders, refunds and locations through your own customer ID; the join to CRM and revenue data with effects and confidence intervals is on Prove at $2,999 a month. Enrichment reads every conversation for its reasons, second reasons, effort and outcome, with custom Dimensions written in plain English from your own Topic hierarchy (custom Dimensions start on Discover at $999 a month). The visual analytics layer builds the weekly contact reasons report, a deck for the operations review and a dashboard the product team can open, downloadable as PDF or PowerPoint, shared by a link you can revoke, and rerun on a schedule into the Final Reports folder.

Dropbox used Dimension Labs to identify its top support intents and the drivers of unresolved conversations, and reports a 43% reduction in unhandled bot conversations. See the Dropbox story.

When Gladly alone is the better choice: if agents apply Topics on nearly every conversation and you need counts and trends, the Topics report and dashboard give you that inside Gladly, and Automatic Topics covers the conversations Gladly's AI handles. Reading every conversation across channels, checking Topics against the text and ranking reasons by what they cost is the job Dimension Labs is built for.

FAQ

Can I export Gladly conversations with the full text?

Yes, through Gladly's API. Its export files arrive every 24 hours by default and stay available for 14 days. Calls come with a recording link and, if you use Voice Summaries, a text summary.

Does Gladly's Topics report count a conversation more than once?

A conversation with several Topics counts under each, and a sub-topic also counts toward every parent above it. Use the CSV's directly applied column to compare reasons at one level.

Should we require a Topic on every Gladly conversation?

Gladly offers the setting and recommends a Topic on every conversation. It improves coverage, and a text check then shows how closely the Topics match what customers wrote.

What is the best way to find top contact reasons across Gladly and other channels?

For a team that wants every Gladly conversation read with the same reason list as its reviews, surveys and other support tools, joined to order data and reported to each owning team, Dimension Labs. For Topic counts inside Gladly, the Topics report is enough. For an Intercom workspace, our post on exporting Intercom conversations for contact reasons covers the same steps.

To see your own Gladly conversations ranked by reason and by cost, start with one export file. Every plan from Explore to Scale starts with a 14-day free trial.

Dimension Labs/Start with one file From $299 a month

Stop reporting. Start understanding.

Start with one file. Ask why a number moved. Plans from $299 a month, no demo and no contract on monthly plans.