Every customer conversation contains signals about the future of the relationship. Some are explicit: "We are evaluating other options." Most are subtle: a shift in tone, fewer questions about the roadmap, a new stakeholder who was not part of the original deal. These signals exist whether or not you catch them. The question is whether you have a system for detecting them.

Most retention programs rely on product usage data, NPS scores, and support ticket volume. These are lagging indicators. By the time usage drops or a customer gives you a 6 on an NPS survey, the decision to churn may already be made. The leading indicators live in conversations, in what customers say during check-ins, QBRs, and project calls.

The Gap Between What Customers Say and What Gets Logged

Sales and customer success teams have always known that conversations contain the best signals. The problem is capturing them at scale. A CSM with 25 accounts has dozens of meetings per month. After each one, they are supposed to update the CRM with relevant signals. In practice, they log a brief summary and move on to the next call.

The result is that retention signals get filtered through two lossy processes: the CSM's ability to notice them in real-time, and the CSM's willingness to log them afterward. Important nuance gets lost at both stages.

AI meeting transcription removes both bottlenecks. Every word is captured, attributed to the right speaker, and available for automated analysis.

What Churn Signals Sound Like

Churn signals are rarely dramatic. Customers do not typically announce "We are leaving." Instead, the language shifts gradually over weeks and months.

Early Signals (3 to 6 months before churn)

Decreasing engagement with your roadmap:

"We are less focused on new features right now"
Questions about upcoming functionality become less frequent
The customer stops asking "When will you support X?"

Subtle ownership language shifts:

Moving from "our tool" or "our platform" to "your tool" or "your product"
Less first-person language about the product: "We use it" becomes "It is being used"

Organizational uncertainty:

"We are going through some restructuring"
"There is a new VP who wants to review all our vendors"
"I might be moving to a different role"

Mid-Stage Signals (1 to 3 months before churn)

Direct cost scrutiny:

"We need to look at where we can cut costs"
"Can you walk me through exactly what we are paying for?"
"Our CFO is asking about the ROI on every tool"

Competitive evaluation:

"We have been getting outreach from [competitor]"
"A colleague at another company uses [competitor] and mentioned it"
"How do you compare to [competitor] on [specific feature]?"

Reduced meeting engagement:

Shorter meetings with fewer questions
Cancellations and reschedules become more frequent
The champion sends a delegate instead of attending personally

Late Signals (less than 1 month before churn)

Contract-specific language:

"When does our contract renew?"
"What is our cancellation process?"
"Can we go month-to-month instead of annual?"

Data extraction requests:

"How do we export our data?"
"Can we get an archive of everything?"

By this stage, the customer has usually made their decision. Intervention is difficult.

Setting Up Automated Churn Detection

Smart Tags let you define detection criteria that run automatically on every meeting transcript. Here is a practical setup for retention monitoring:

Tag: "Churn Risk, Budget"

Detection criteria: budget cuts, cost reduction, ROI question, too expensive, value for money, spending review, vendor consolidation, cut costs, reduce spend

Tag: "Churn Risk, Competition"

Detection criteria: evaluating alternatives, looking at other options, competitor names, switching to, compared to, heard about, demo with

Tag: "Churn Risk, Engagement"

Detection criteria: not using, underutilized, have not had time, too busy to implement, sitting on the shelf, gathering dust

Tag: "Churn Risk, Organizational Change"

Detection criteria: restructuring, new leadership, reorg, change in direction, shifting priorities, new VP, new CTO, strategic review

Tag: "Churn Risk, Contract"

Detection criteria: cancellation, cancel, end the contract, not renew, month to month, downgrade, reduce seats

When any of these tags fire after a customer meeting, the account team gets immediate visibility into what was said and by whom. This is not sentiment analysis guessing at tone. It is specific language from the transcript, attributed to a named speaker, with the surrounding context.

What Expansion Signals Sound Like

Expansion signals are the other side of the coin. These are the moments when a customer reveals that they are getting value and could use more.

Usage Expansion

"We have been using it more than we expected"
"Our team in [other office/department] has been asking about it"
"Can we add more seats?"
"We are onboarding the new hires and they all need access"

Feature/Tier Expansion

"Do you have anything for [adjacent use case]?"
"What would we get if we upgraded to the next tier?"
"We have been working around [limitation] but it would be great to have [premium feature]"
"Is there an API we can use to integrate this with our other tools?"

Strategic Commitment

"We want to make this the standard across the company"
"Can we do a multi-year deal?"
"This has become part of our core workflow"
"Our VP mentioned this in the all-hands as a win"

Smart Tags for Expansion

Tag: "Expansion, New Users" Detection criteria: add seats, more licenses, new team, new department, new office, onboarding, roll out, scale up, hiring

Tag: "Expansion, Feature Interest" Detection criteria: upgrade, premium, advanced features, enterprise tier, API access, integration, additional modules, more capabilities

Tag: "Expansion, Strategic Value" Detection criteria: standard across, core workflow, company-wide, essential tool, multi-year, long-term, committed to

Building a Signal-Based Retention Program

Individual signal detection is useful. The real impact comes from aggregating signals across accounts and over time.

Weekly Review Process

Review new Smart Tag alerts. Which accounts triggered churn risk tags this week? Which triggered expansion tags?
Cross-reference with account data. Does the churn signal match other indicators (declining usage, open support tickets, upcoming renewal)?
Prioritize intervention. Not every churn signal requires immediate action. A single mention of "budget review" is worth noting. Three budget-related signals in two months is a pattern that needs a response.
Assign action. For churn risk accounts, schedule a value review meeting. For expansion signals, loop in the sales team to scope the opportunity.

Using AI Chat for Deeper Analysis

When a Smart Tag fires and you want more context, use the AI Chat feature to query across recent meetings with that account:

"Has [customer] mentioned budget concerns in any meetings over the last quarter?"
"What has [customer] said about their experience with our product in the last three calls?"
"Are there any unresolved issues or commitments we have not followed through on with [customer]?"

This cross-meeting analysis often reveals that a single signal is part of a larger pattern, or alternatively, that it was a one-time comment that does not indicate a trend.

CRM Integration

When IceCubes syncs meeting data to HubSpot or Salesforce, Smart Tag results flow alongside the standard meeting summary. Your CRM records show not just "meeting happened" but "this meeting contained expansion signals related to new department interest." Over time, this creates a data layer in your CRM that supports more nuanced account health scoring.

Measuring Impact

Track these metrics before and after implementing signal-based retention:

Time to detect churn risk: How many days before renewal do you first identify an at-risk account?
Intervention success rate: Of accounts flagged as at-risk where you intervened, what percentage renewed?
Expansion capture rate: Of accounts that showed expansion signals, what percentage converted to an upsell?
Surprise churn rate: What percentage of churned accounts were not flagged in advance?

The goal is to shift from reactive retention (responding after a customer says they want to cancel) to proactive retention (addressing concerns before the customer reaches that point).

Getting Started

Install IceCubes and set up the churn risk and expansion Smart Tags described in this post. Run them for a month across your customer-facing meetings. Review the signals weekly. You will likely find that your meetings were telling you more than you realized.

50 free AI credits, no credit card required.

Add to Chrome | Add to Edge

For more on Smart Tags, see our full guide: Smart Tags for Sales Teams. For customer success-specific use cases, see Meeting Transcription for Customer Success Teams.