What You'll See in the Dashboard
The dashboard surfaces churn risk and lifetime value per visitor — churn as a probability score (0–100) and LTV as a 0–100 score with a tier and confidence. Filter visitors by churn risk to see which customers need attention right now. (Behavioral scoring models are available on Growth plans and above.)
Business Actions: Trigger a retention offer when churn risk climbs. Pipe high-LTV visitors into your CRM for white-glove outreach. Segment audiences by LTV tier to optimize ad spend on your most valuable cohorts.
Model 10: Churn Prediction
The churn prediction model estimates the probability that a visitor or customer will not return. For SaaS products, this predicts subscription cancellation; for e-commerce, it predicts the end of a customer relationship. ClickStream detects churn signals during active sessions, enabling intervention while the user is still present.
The 9 Churn Signal Families (Illustrative)
The signal families below illustrate the kinds of on-site behavior the model weighs. The exact signal list and weights are internal — and tunable per site via scoring configuration.
| Signal | Description |
|---|---|
| Declining session frequency | Interval between sessions is increasing over the last 4 visits |
| Declining engagement depth | Engagement score trending downward across sessions |
| Reduced feature usage | Using fewer product features per session than historical average |
| Support/cancel page visits | Viewing cancellation, downgrade, or complaint pages |
| Competitor comparison signals | Referrer from competitor review sites; comparison page visits |
| Declining session duration | Sessions getting shorter over time |
| Error-frustration accumulation | Frustration score increasing across multiple sessions |
| Billing page pattern | Visiting billing settings without making a purchase |
| Content disengagement | Reduced return-visit depth and declining on-site content interaction across sessions |
The 4 Churn Levels (Suggested Operating Framework)
The model’s output is a 0–100 churn risk score. The bands and level names below are a suggested operating framework for acting on that score — not labels the product emits.
| Score Range | Level | Description | Intervention Urgency |
|---|---|---|---|
| 0–25 | Healthy | No churn indicators present. User is engaged and active. | None. Continue normal experience. |
| 26–50 | At Risk | Early warning signals detected. Engagement declining. | Low. Proactive engagement nudges. |
| 51–75 | Likely Churning | Multiple strong churn signals. Clear disengagement pattern. | Medium. Retention offer or outreach. |
| 76–100 | Imminent Churn | Active churn behavior (cancel page, billing review). | High. Immediate intervention required. |
Churn Risk Factors
SaaS vs. E-Commerce Churn Signals
| Signal Category | SaaS Indicator | E-Commerce Indicator |
|---|---|---|
| Usage decline | Fewer logins, less feature usage, shorter sessions | Less frequent visits, smaller browsing sessions |
| Active disengagement | Export data, cancel page, downgrade page | Unsubscribe page, email opt-out, account deletion |
| Competitor signals | Referrer from G2/Capterra competitor pages | Referrer from competitor product pages |
| Support pattern | Increasing support tickets with declining satisfaction | Complaint page visits, return policy views |
| Financial signals | Billing page views, plan comparison without upgrade | Coupon-only purchasing, declining AOV |
Model 11: LTV Score
The LTV score predicts the total value a visitor will generate over their entire customer relationship. Unlike the value estimator (Part 2) which focuses on the current session, LTV scoring projects forward across months or years.
The 8 LTV Signal Families (Illustrative)
As with churn, these signal families illustrate what the model weighs; the exact signal list and weights are internal and tunable per site.
| Signal | Description |
|---|---|
| Purchase frequency trend | Accelerating purchase frequency = higher LTV |
| Average order value trend | Increasing AOV over time = growing relationship |
| Engagement depth consistency | Sustained high engagement = loyal customer |
| Product category breadth | Cross-category purchasing = deeper relationship |
| Referral source quality | Organic/direct visitors have higher average LTV |
| Session recency | Recent high-engagement sessions boost LTV |
| Feature adoption velocity | Speed of adopting new features (SaaS-specific) |
| Social engagement | Reviews, referrals, community participation |
The 4 LTV Tiers (Suggested Operating Framework)
The shipped model outputs a 0–100 LTV score with a tier and a confidence value. The tier names and bands below are a suggested operating framework, not documented product output.
| Score Range | Tier | Description | Strategy |
|---|---|---|---|
| 0–25 | Low LTV | One-time or infrequent buyer. Limited growth potential. | Efficient self-service. Minimize support cost. |
| 26–50 | Medium LTV | Repeat buyer with growth potential. | Cross-sell, loyalty programs, engagement campaigns. |
| 51–75 | High LTV | Loyal customer with consistent purchasing. | Priority support, exclusive offers, early access. |
| 76–100 | Whale | Top-tier customer. Significant revenue contribution. | Dedicated account management, VIP treatment. |
Real-Time vs. Historical LTV
ClickStream computes two LTV variants:
- Session LTV (real-time): Projected lifetime value based on behavioral signals observed in the current session. Available for both new and returning visitors.
- Historical LTV: Cumulative LTV based on actual purchase history and behavioral trends. Only available for returning visitors with transaction data.
One illustrative way to blend the two: ltvScore = 0.6 * historicalLTV + 0.4 * sessionLTV for returning visitors, and ltvScore = sessionLTV for new visitors. The shipped model’s exact weighting is internal and tunable per site.
Churn × LTV Priority Matrix
The intersection of churn risk and LTV creates a prioritization framework for customer success teams:
| Low LTV (0–25) | Medium LTV (26–50) | High LTV (51–75) | Whale (76–100) | |
|---|---|---|---|---|
| Healthy (0–25) | Auto-nurture | Cross-sell campaigns | Loyalty program | VIP monitoring |
| At Risk (26–50) | Automated re-engagement | Targeted email + incentive | CSM outreach | Executive escalation |
| Likely Churning (51–75) | Win-back campaign | Retention offer | Personal call + special offer | C-level intervention |
| Imminent (76–100) | Exit survey | Last-chance offer | Urgent 1:1 save attempt | All hands on deck |
Real-Time Intervention Timeline
Here is an example of how churn signals evolve during a single session for a SaaS customer who is considering cancellation:
| Time | Action | Churn Score | LTV Score | System Response |
|---|---|---|---|---|
| 0:00 | Login to dashboard | 35 | 72 | Monitor: at-risk + high-LTV flagged |
| 0:30 | Navigates to billing settings | 52 | 72 | Alert: CSM notified via your webhook → Slack workflow |
| 1:15 | Views plan comparison page | 58 | 72 | Show: "Need help choosing a plan?" chat widget |
| 2:00 | Clicks "Cancel Subscription" | 85 | 72 | Intercept: retention offer + personal call scheduling |
| 2:30 | Views retention offer | 78 | 72 | Monitor: did they accept? |
| 3:00 | Accepts offer, returns to dashboard | 45 | 72 | Success: tag for follow-up in 7 days |
Alert Integration
Churn and LTV alerts can reach your existing tools via ClickStream’s signed webhook destination:
The churn-LTV matrix transforms retention from a reactive process (noticing a cancellation after it happens) into a proactive one (intervening during the session where churn behavior first appears). For high-LTV customers, this single capability can justify the entire ClickStream investment.