How to spot churn signals in Help Scout
Find churn signals in Help Scout: AI Topics, tags, custom fields, customer profiles, views, the All Channels report and the Inbox API, step by step.
By Florian
Help Scout is built to make each conversation feel personal, and it does that well. The side effect is that patterns across conversations are easy to miss: the customer who has written four times this quarter, the rise in “can I get a cheaper plan?”, the account that asked how to export everything two weeks before cancelling.
This guide shows how to make those patterns visible in Help Scout: what to label, how to label it consistently, where to look, and how to get the data out to check it against real cancellations. It relies on Help Scout’s own documentation: AI Topics, categorizing incoming email, tags, custom fields and reporting on them, customer profiles, views, reports, data export options and the Inbox API (all accessed November 2026). For satisfaction ratings specifically, see our separate guide on Help Scout ratings as a churn signal.
Three kinds of signal
Churn signals in a help desk come in three kinds, and Help Scout handles each differently:
| Kind | Example | Best Help Scout tool |
|---|---|---|
| What the customer writes about | Pricing, export, cancellation, a competitor | Tags, AI Topics, custom fields |
| How often they write | Third conversation about the same bug | Customer profile, Conversations card |
| How it went | Long wait, several replies, unresolved | Reports, satisfaction ratings |
Most teams only track the first. The second is often the strongest, and it is the one Help Scout makes easiest to see for a single customer and hardest to count across all of them.
Labelling what customers write about
Tags from workflows
Help Scout’s categorizing guide describes adding tags from workflows based on keywords in the subject or body, on how long an email has waited for a reply, or on the sender. For churn, a small set of tags is enough:
churn-cancel: cancel, close account, unsubscribechurn-price: too expensive, cheaper plan, discountchurn-export: export, download all, delete my datachurn-competitor: names of tools you lose customers to
Keyword tags are noisy. Ask agents to remove wrong ones; a tag nobody trusts is worse than no tag.
AI Topics
Help Scout also offers Topics, currently in beta. Per Help Scout, when you enable it for an inbox, the AI proposes roughly 8 to 15 topics based on your history, which you review before approving; conversations it cannot classify go to an Uncategorized topic. During the beta, manual overrides do not feed back into the model, and Topics can be used as a condition in workflows but not as an action.
Topics and your own tags answer the same question from different directions. Topics find themes you did not think of; tags track the specific signals you decided matter. Use both, and compare them once a month.
Custom fields for the reason
Custom fields are available on the Plus and Pro plans; a dropdown can have up to 100 options. A single dropdown “Risk reason” (Price, Missing feature, Bug, Support experience, Switching, None) set by the agent when closing a conversation gives cleaner data than tags. Help Scout can report on custom fields in the All Channels report, showing each option’s share and the matching conversations.
Seeing how often a customer writes
Help Scout creates a customer profile automatically when someone first writes in, and the profile shows their full conversation history. On Plus and Pro, the sidebar’s Conversations card combines recent conversations from the contact and their company (company management).
That is the right place for an agent to notice “this is the third time”. Make it a habit: before replying to a customer with several recent conversations, read the last two. If the same problem keeps coming back, say so in the reply and escalate it.
Counting repeat contacts across all customers is not something the standard reports show directly. That is a job for the API (below).
Where to look each week
Help Scout views filter conversations by conditions including tags (“is tagged with any”, “is tagged with all”, “is not tagged with”). Create one view per churn tag, and one combined view “Churn signals: last 7 days”.
The All Channels report shows how often each tag is used and whether it is rising or falling, and report views can filter report data by inbox, tag, folder, custom field or conversation type. A rising churn-price line over two months is worth a conversation with whoever owns pricing.
A weekly routine:
- Open the “Churn signals: last 7 days” view and read each conversation, not just the subject.
- Fix wrong tags and set the “Risk reason” field.
- For each customer with a signal, check their profile for earlier conversations.
- List customers worth a personal follow-up: high value, several signals, or an unresolved problem.
Checking signals against real churn
Tags tell you what customers talk about. Only your billing data tells you who left. To connect the two you need conversation data outside Help Scout.
Help Scout’s reporting exports (CSV or XLSX, for selected reports) contain data about each conversation but not the thread contents. For full conversations, Help Scout’s export options point to the API. The List Conversations endpoint returns conversations sorted by createdAt by default, and only active conversations unless you pass status=all. List Threads returns each message with its type and createdAt.
With that data:
- Join conversations to subscriptions by customer email.
- For each customer who cancelled, look at the 90 days before: which tags, how many conversations, any unresolved ones.
- Do the same for a similar group who did not cancel.
- Keep the signals that show up clearly more often before cancellations. Drop the rest.
Report counts with every percentage. A tag that appears before 4 cancellations out of 6 is a lead to investigate, not a finding. Our guide on finding what predicts churn covers the method.
What to do with a signal
A churn signal is a reason to help, not a reason to sell:
- Price: tell the customer about smaller plans or annual pricing if they exist. Do not invent a discount.
- Bug, again: escalate, and tell the customer what you did.
- Export: answer fully and quickly. Making export hard does not retain anyone.
- Cancel: cancel. See our guide to handling cancellation requests in Help Scout.
Doing this with NoChurn
NoChurn’s Help Scout connector is read-only: it reads conversations and never replies, tags or edits anything in Help Scout. Conversations are matched to your customers by email; doubtful matches are suggested for you to confirm and never merged automatically. NoChurn combines them with billing data in a churn score from 0 to 100 and a churn drivers view that ranks attributes by correlation with churn, and keeps messages as dated, verbatim evidence for disputes on Stripe and Braintree. Tags, Topics, views and replies stay in Help Scout. See NoChurn + Help Scout.