Skip to main content

Churn Prediction

Last updated: 2026-05-18

SaaSy's churn prediction engine runs independently from the health score. Where the health score is a snapshot, churn prediction is forward-looking: it weighs behavioral trajectories to estimate the probability that a customer will not renew.

How probability is calculated

The engine adds risk contributions from several signal groups, then normalises to a 0.0–1.0 probability. The probability maps to a risk level:

ProbabilityRisk level
< 0.30Low
0.30–0.49Medium
0.50–0.69High
≥ 0.70Critical

Engagement signals

SignalContribution
Days since last login > 30+0.25
Days since last login 15–30+0.15
Active user percentage < 30%+0.20
Feature adoption rate < 40%+0.15
SignalContribution
Login frequency trend: declining+0.20
Feature usage trend: declining+0.20
Payment issues (per incident, max 3)+0.15 each

NPS

NPS scoreContribution
0–6 (Detractor)+0.25
7–8 (Passive)+0.10
9–10 (Promoter)-0.00 (protective — see below)

Contract and account factors

SignalContribution
Auto-renew off AND renewal in < 60 days+0.15
More than 2 support escalations in 30 days+0.20
Any downgrades in account history+0.15

Protective factors

These factors don't reduce the raw score directly, but they are surfaced in the prediction output and used to qualify the confidence level:

  • Auto-renewal enabled — reduces urgency of intervention.
  • Feature adoption rate > 60% — indicates product stickiness.
  • NPS ≥ 9 (Promoter) — strong satisfaction and loyalty signal.
  • Contract value > $100k — enterprise contracts often have structural retention.

Confidence score

The prediction comes with a confidence level (0.0–1.0) based on how many signals were available:

Signals availableConfidence
5 or more0.85–0.90
3–40.75–0.82
1–20.65–0.72

When SaaSy has access to your live database (all connected integrations), confidence increases because it can pull real activity data rather than relying on interpolated inputs.

Risk factors and interventions

For each prediction, SaaSy surfaces the top contributing risk factors and recommends interventions based on the risk level:

Risk levelRecommended interventionTimeline
CriticalExecutive escalation + retention packageWithin 48 hours
HighSuccess plan resetWithin 1 week
MediumHealth check callWithin 2 weeks
LowRegular touchpointMonthly cadence

Factor-specific interventions are also generated when identifiable triggers exist:

  • Low User Adoption → Targeted activation and training campaign.
  • Payment Issues → Billing review and payment terms conversation.

Trend analysis

The prediction output includes a trajectory based on the previous prediction:

Change in probabilityTrajectory
+15% or moreRapidly Deteriorating
+1% to +14%Worsening
-1% to -14%Improving
-15% or moreImproving Significantly
Within ±1%Stable

Momentum is also tracked:

  • Negative — both login frequency and feature usage trending down.
  • Positive — both trending up.
  • Mixed — trends diverging.

How alerts fire

A RenewalRisk alert fires when churn probability crosses the threshold configured in Settings > Alert Rules. By default this is set to fire at High risk level (probability ≥ 0.50). You can lower this to Medium (0.30) to get earlier warnings at the cost of more noise.

See Alerts & Rules for how to configure thresholds and notification channels.

Viewing predictions

Open any customer from the Customers dashboard and click the Churn Risk badge. The detail panel shows:

  • Current probability and risk level.
  • Top risk factors with severity and evidence.
  • Protective factors.
  • Recommended interventions with assigned owner and timeline.
  • Trend trajectory vs. previous prediction.