Why churn is hard to predict

Every team sees part of the picture. None sees the whole thing.

Customer Success watches health scores and NPS responses. They notice when engagement drops in their QBRs—but that's quarterly, and churn doesn't wait.

Support sees ticket volume and sentiment. They know when customers are frustrated—but they don't see usage data or payment history.

Product tracks feature adoption and activity metrics. They see usage decline—but they don't know if the customer is unhappy or just on vacation.

Finance catches payment failures and downgrades. By then, the customer has already decided to leave.

The signals that predict churn are distributed across all these systems. No single team has the complete view.

Signals that indicate churn risk

Research on SaaS churn consistently identifies these patterns:

Usage decline

Active users dropping. Login frequency decreasing. Core features unused. This is the most reliable signal—but it lives in product analytics, not CRM.

Support patterns

Increasing ticket volume. Negative sentiment in conversations. Escalations to management. Unresolved issues that linger. This lives in Intercom or Zendesk.

Payment friction

Failed charges. Expired cards not updated. Requests for invoicing changes. Delayed payments. This lives in Stripe.

Engagement decay

Emails unopened. QBRs declined. Champion leaving the company. Contract renewal conversations going silent. This lives in the CRM.

The compound signal

Any one of these might mean nothing. Together, they tell a story. A customer with declining usage AND increasing support tickets AND a failed payment is almost certainly at risk. But seeing that requires data from three different systems.

How Eru identifies at-risk accounts

Unified customer view

Eru maps customer entities across all your systems. The account in your database, the company in HubSpot, the customer in Stripe, the workspace in Intercom—they're linked. When you look at a customer, you see everything.

Cross-system watchlists

Eru creates watchlists that analyze signals from multiple sources. Not a simple filter on one metric—a scored list based on conditions across your entire data stack.

Watchlist: At-Risk Accounts Entity: customer Conditions: - Usage: MAU dropped >30% vs 90-day average (from product database) - Support: >3 tickets in last 30 days OR negative sentiment detected (from Intercom) - Payment: Any failed charge in last 60 days (from Stripe) - Engagement: No login by primary contact in 14+ days (from product database) Scoring: - Each condition met: +25 points - 100 points = critical risk Output: - Customer name and tier - Account owner (from HubSpot) - MRR at risk (from Stripe) - Top signals triggered - Deep links to each source system Alert: When list grows by >10% or when Enterprise account appears

Evidence-backed insights

When an account appears on the watchlist, you see exactly why. Which signals triggered. What the actual data shows. Links to investigate in each source system. Not a black-box score—transparent analysis you can act on.

Continuous monitoring

Watchlists run on a schedule. Daily, hourly, whatever makes sense for your business. When the list changes significantly—new accounts added, accounts moving to higher risk—you get alerted in Slack.

From detection to action

Knowing who's at risk is only useful if you can act on it.

Prioritized outreach

The watchlist is sorted by MRR at risk, not just risk score. A $50/month account at critical risk matters less than a $5,000/month account at moderate risk. Focus where it matters.

Context for conversations

When CSMs reach out, they know why. "I noticed your team's usage dropped after your lead developer left. How can we help with the transition?" is more effective than "Just checking in!"

Pattern recognition

Over time, you see which signals actually predict churn in your business. Maybe support tickets matter more than usage. Maybe payment failures are a lagging indicator. Eru's evidence packs let you analyze what actually correlates with churn for your customers.

Beyond churn: other watchlists

The same cross-system analysis works for other use cases:

Related

Stop reacting to churn

Eru surfaces at-risk accounts before they become cancellation emails.