ANSWER

How to Detect Churn Risk From What Customers Say in Tickets and Calls

Customers rarely say “churn.” They say “this is the third time I’ve asked about this.”

Try Rippit free
or request demo
By Lauren Alexander · Published October 2, 2026

You can detect churn risk from conversations by tagging every ticket and call for five signal families, scoring each for severity and persistence, and flagging accounts where an unresolved signal keeps coming back. With Rippit, an agent you set up reads every ticket, chat and call you connect next to CRM, renewal and adoption data, and flags the accounts where risk is building.

Picture an account a few months from renewal. On a routine CSM call, your champion mentions she’s moving to a new role and that “Sam will own this now.” Usage doesn’t move, the health score stays green, and nobody connects the handover to the two unresolved tickets her team opened last month. What customers say is often the first warning that an account is slipping.

Churn risk isn’t a sentiment problem. It’s a pattern-recognition problem.

Key takeaways

  • Five signal families cover most churn language.
  • One angry, resolved ticket is noise. The same unresolved problem raised three times is a pattern.
  • A champion leaving happens once, so it won’t build persistence. Confirm it with CRM and people data and act within days, instead of waiting for it to repeat.
  • Score severity and persistence, then calibrate the threshold against your own renewals and churn.
  • If 2% of conversations contain a signal and you review a random 2%, you read about 1 in 50 of the conversations that contain it.

What churn signals should you look for in customer conversations?

Start with five signal families.

They aren’t keyword searches: “cancel” might mean a duplicate invoice.

Five churn signal families in customer conversations
Signal familyWhat it might sound likeWhy it matters
Repeated unresolved friction“This is the third ticket,” “We’re still having the same issue”The problem persists despite repeated fixes
Escalation / executive involvement“I’m looping in our VP,” “Can someone senior call me?”The issue is moving beyond the day-to-day user
Alternative evaluation“We’re looking at other options,” “Another vendor supports this”The customer is considering a different solution
Value / commercial pressure“What are we paying for?” “We aren’t using enough of this”The customer is questioning the economics
Relationship disruption (champion change)“I’m leaving the company,” “Sam will own this now”An internal advocate may be changing

Support usually hears these first, in tickets and chats, often before the CSM does; CSM calls and account reviews carry the rest. Read them together, per account, so a signal in one channel isn’t lost because the others look quiet.

How do you tell frustration from churn risk?

Separate severity (how consequential is the signal?), persistence (is it recurring?) and resolution (was the problem actually fixed?).

An angry customer whose ticket is fixed in 15 minutes is fine. A polite one who reports the same integration failure three times in a month, then says “we’re reviewing alternatives before renewal,” deserves the attention.

How do you score churn risk without drowning CSMs in alerts?

Use a starting rubric, then calibrate it.

Score each signal family 0–3 on severity and 0–3 on persistence, and take the highest of each for the account.

Starting churn-risk rubric: severity and persistence, scored 0–3
ScoreSeverity (per conversation)Persistence (per account, trailing 60 days)
0Signal absentSeen once, resolved
1Present, mild, resolved in-conversationSeen twice, resolved each time
2Present, unresolved at closeSeen 3+ times, or twice unresolved
3Unresolved, with authority or competitor languageSeen 3+ times unresolved across 3+ weeks

Account risk = highest severity × highest persistence. As a starting point, 6 or above is a flag; 3 to 5 goes on a watchlist and decays to zero after 60 quiet days, so one bad Tuesday doesn’t haunt an account for a year.

These numbers aren’t universal. Look back at the conversations of accounts that churned and accounts that renewed, and adjust until the flags separate the two. AI can find the signals. Your data should determine which signals actually matter.

Then rank the flagged accounts by revenue at risk, using ARR and renewal date from your CRM: the same score matters more on a large account renewing next month than on a small one renewing next year. Our playbook on how to build a churn radar covers the account roll-up, the backtest and the schedule.

How do you tell a real signal from noise in a single conversation?

Read each line for what it is, not how it sounds. Here is one CSM call with the champion from the opening, five lines worth a second look:

Reading one CSM call: which lines are churn signals, and what to check next
What the customer saidSignal familyReadCheck next
“The export bug drove me nuts last week, but support fixed it fast.”Repeated unresolved frictionNoiseResolved, seen once. Leave it unless it comes back.
“I’m moving to the ops team next month. Sam will own this now.”Relationship disruption (champion change)SignalContact roles in the CRM; a job change in people data; whether Sam has ever joined a call
“Sam’s team hasn’t really used the reporting side yet.”Value / commercial pressureWeak aloneActive users and feature adoption for Sam’s team. Falling usage turns it into a signal.
“Every tool is getting reviewed before the new VP’s first planning cycle.”Escalation / executive involvementSignalThe renewal date and whether the opportunity has a next step; the new VP’s name and start date
“We used another vendor at my last company.”Alternative evaluationNoiseA past mention, not an evaluation. Watch for a demo, a trial or a price-match request.

Three tests sort most lines: was it resolved in the conversation, is it about the future or the past, and does it change who decides the renewal? The last test matters most here. A champion change happens once, so it never builds persistence, and the rubric alone would leave this account on the watchlist. Treat it as an event instead; the first 48 hours, below, show what to do.

What does a churn pattern look like over time?

Take a hypothetical account. Day 1: a data sync fails and is resolved (severity 1). Day 12: it fails again, “This is the second time and it cost us a reporting cycle” (severity 2). Day 21: the user’s manager asks for a remediation plan and says the team is “looking at what else is out there” (severity 3).

1 · SCAN · Every conversation on the account

Day 1
Severity 1 · resolved

Data sync fails; fixed in the conversation.

Day 12
Severity 2 · unresolved

“This is the second time and it cost us a reporting cycle.”

Day 21
Severity 3 · unresolved

Manager asks for a remediation plan; team is “looking at what else is out there.”

2 · ANALYZE

Score the sequence, not one ticket

Highest severity 3 · Persistence 2 · Three occurrences, two unresolved

Account risk 3 × 2 = 6

3 · ACT

Flag the account, with evidence

Flag: 6 or above

The CSM gets the three conversations behind the score, not just a number.

No single conversation tells the story. The sequence does. Three occurrences, two unresolved, is persistence 2, so the account scores 3 × 2 = 6 and is flagged. That’s more useful than “Sentiment: Negative.”

Why should you analyze every conversation, not a sample?

Because churn signals are sparse.

If 2% of conversations contain a signal and you review a random 2%, you read about 1 in 50 of the conversations that contain it; the other 49 go unread. Klaviyo handles 600,000 to 700,000 support incidents a year, and surveys came back on only about 5% of emails and 20 to 25% of chats (Klaviyo story).

Brex now checks every support ticket for churn risk and finds it in 2 to 3% of tickets each week. “With CSAT, you have to wait for something to go wrong and for somebody to say something before you reach out. With these cases, we’re getting ahead of it. Within hours, somebody is checking in with the customer, maybe before they’ve even had a chance to submit a survey,” says Medha Mayawala, Business Development at Brex.

Use samples to validate the analysis. Use the full dataset to measure the business. Humans check that your signal definitions work; the validated definitions then run across every relevant conversation (why Rippit reads 100% of conversations).

What should happen after a churn signal is detected?

The right first move depends on the signal.

The first question to ask for each churn signal
SignalUseful first question
Repeated unresolved frictionHas the underlying problem actually been fixed?
Executive escalationWho internally should own the response?
Alternative evaluationWhat problem is the customer trying to solve with an alternative?
Value / commercial pressureWhat value does the customer believe they’re not receiving?
Relationship disruptionWho owns the relationship now, and do they understand the value delivered?

For alternative evaluation specifically, our playbook on tracking competitor mentions across sales and support conversations shows how to separate real migration threats from passing references.

A CSM shouldn’t receive “Acme Corp—Churn Risk: 87.” They should receive “Acme has raised the same data-sync problem three times in 21 days; a manager joined and mentioned alternatives,” with the conversations behind it.

NOT THIS

Acme Corp

Churn Risk 87

No reason, no conversations. The CSM has to start from scratch.

THIS · RIPPIT DAILY CHURN RISK ALERT

#cs-churn-alerts · 8:00 AM

Acme has raised the same data-sync problem three times in 21 days; a manager joined and mentioned alternatives.

Repeated unresolved frictionEscalationAlternative evaluation
Risk 6 · severity 3 × persistence 2

Ticket #31842 · Day 12 “This is the second time and it cost us a reporting cycle.”
Call · Day 21 Team is “looking at what else is out there.”

First question: Has the underlying problem actually been fixed?

A flag a CSM can act on names the pattern, the signals and the conversations behind it, plus the first question to ask.

A flag without evidence is just another score.

What should you do in the first 48 hours after a champion-change flag?

Move while the departing champion is still reachable and before the new owner forms a view of you. For the account in the opening, that means reaching Sam before the renewal conversation starts:

  1. 1

    Confirm

    Check contact roles in the CRM, job-change data and the call where it was said.

  2. 2

    Count the threads

    Who else from the account shows up in tickets and calls? Single-threaded accounts go first.

  3. 3

    Ask for the handover

    An intro from the departing champion to the new owner, while they’re still there.

  4. 4

    Brief the new owner

    Value delivered and open issues, taken from the conversations rather than memory.

  5. 5

    Result

    CRM contacts and roles updated, and the account watched for 60 days.

The handover is the key step: it’s the one you can’t do once the champion has gone.

Two mistakes to avoid. Don’t make the renewal the first thing you raise with Sam; start with the two unresolved tickets and what you’re doing about them, because those are what Sam will hear about first. And don’t wait for the CRM to catch up. 76% of CRM users say less than half of their company’s CRM data is accurate and complete (Validity, 2025), so the conversation where the champion said goodbye is often the only record that anything changed.

How does Rippit detect churn signals?

Rippit treats churn detection as one analysis on the same customer and business data used for insights, QA and AI-agent monitoring. Connect your conversation tools with pre-built integrations, including Zendesk, Intercom and Gong, and your CRM and product data, such as Salesforce and HubSpot, through our MCP where supported, and define the signal in natural language: “Identify conversations where a customer describes a recurring unresolved product problem. Do not flag a single resolved incident.”

Rippit applies that definition to every conversation you connect, and to new ones as they sync. Each result becomes structured fields (signal family, severity, resolved, account, date, evidence), saved next to the account’s ARR, renewal date and product adoption, so the rubric above runs on recorded data rather than memory. An agent you set up rolls them up by account and delivers flagged accounts on a schedule to Slack or email, or you ask Claude, ChatGPT or Cursor through Rippit’s hosted MCP server. Through MCP, the agent can also read data from the rest of your stack, such as Snowflake.

SpotOn had Rippit read two years of churned accounts across Salesforce, Gong and Zoom, and now a risk signal opens a Salesforce case for a success manager within one hour (SpotOn story). “Now when a risk signal goes off in Rippit, there’s a case automatically generated in Salesforce for our success managers to proactively engage with that client without them having to ask for it… that’s been a game changer,” says Charlie Caplinger, Program Manager, Voice of the Client. More in the guide to conversation analytics.

Comparing tools for this? See how Rippit stacks up against Gong on sales calls, Qualtrics on surveys and CSAT, and Zendesk on support conversations.

Which agents watch for each churn signal?

One alert that tries to catch all five families tends to catch none of them well. Each family has its own sources, owner and pace, so give each one its own agent on the same joined data:

Five churn signal families and the agent that watches each one
Signal familyAgentWhat it addsRunsSends to
Repeated unresolved frictionChurn-Risk Bug Watch · Built exampleBugs hitting accounts with upcoming renewals (for example, the next 90 days), ranked, with the exact conversationsWeekly by defaultA ranked list to product and engineering in Slack or email
Escalation / executive involvementA custom escalation-watch agent · Example you can buildTickets likely to reach an executive, from tone, repeat contacts, ARR and renewal date, before they doDailySlack to the CSM and support lead
Alternative evaluationCompetitor Threat Alert · Built exampleCustomers who mention a rival, matched to that rival’s latest pricing or launch, with at-risk ARRDaily by defaultSlack to CS leadership, with a talk track for the CSM
Value / commercial pressureRenewal Risk Watch · Built exampleRenewals in a window you set (for example, the next 180 days) whose risk changed, with the reasons and a quoteMondays by defaultThe account’s Slack channel, tagging the CSM, plus a leadership roll-up
Relationship disruption (champion change)A custom champion-change agent · Example you can buildAccounts where the main contact went quiet or left, or a new stakeholder appearedWeeklySlack to the CSM

For Support, Predictive CSAT (Built example) scores every conversation you connect, surveyed or not, and posts a daily low-score list with the driver behind each one.

FAQ: detecting churn risk from customer conversations

Can customer conversations actually predict churn?

Treat them as evidence, not a guaranteed prediction: a customer can show every risk signal and still renew. Backtest the signals against your actual renewal and churn outcomes.

How far before churn do these signals appear?

There isn’t a universal answer; it depends on your business model, renewal cycle and signal type. Measure when signals first appeared before each churn event to get a lead-time distribution for your business.

How is this different from a customer health score?

Logins, seats and feature adoption show what the account is doing; conversations show what the customer is saying. An account can keep heavy usage while building a case to replace you. Behavior tells you what changed. Conversations can help explain why. For the full comparison, see conversation-based churn detection vs. customer health scores.

Won’t this flag every frustrated customer?

It shouldn’t. Detecting negative sentiment will. Checking whether the issue was resolved, has recurred, or has drawn in someone senior detects patterns, not just emotion.

How do you spot a champion change in conversations?

Listen for handover language (“Sam will own this now”), a main contact who stops replying, and new names on calls or copied on tickets. Then confirm it with contact roles in your CRM and job-change data. A contact out for a few weeks with a named backup usually isn’t one.

Do you need a data team to do this?

Not necessarily. With Rippit, business users define analyses in natural language, without needing an engineer to start, but your business experts still need to define what matters.

The one signal that never builds a pattern is the champion who leaves. It happens once, so it needs its own watch.

Example agent you can build · runs weekly

A custom champion-change agent

Catch the handover before the renewal does. Every week it reads new support conversations and sales and CS calls you connect, next to the contact roles in your CRM, and sends:

  • Accounts where the main contact went quiet or left, with the conversation where it showed
  • New stakeholders who appeared on calls or tickets, for the CSM to add as contacts in the CRM
  • Single-threaded accounts, so you can build a second relationship before you need one
  • The ARR and renewal date behind each change, so the riskiest handovers come first
Sends to
SlackEmail

What is Rippit?

The conversation that warns a CSM about an account is also the record that tells Product which bug to fix first and leadership which renewals are at risk.

Rippit is a Growth Data Platform: it keeps enriching your customer conversations and other sources into structured data, connects them to CRM, renewal and product data and each team’s own sources, and lets anyone set up agents on it, without needing an engineer to start. Connect with pre-built integrations or through our MCP, and ask from Rippit, Claude, ChatGPT or Cursor.

Support & CSM conversations (tickets, call transcripts)CRM & renewalsProduct adoptionStakeholder engagement (champion changes)Competitor & market shiftsand more

Build your first agent on your customer and business data.

Try Rippit free
or request demo

Sources

  • Validity, The State of CRM Data Management in 2025

Keep reading

More from the Rippit Library.

View the library →