Several MagicAI Engage trigger rules target members by churn risk. This article explains where the churn signal comes from, the five risk bands, how new joiners are assessed by dedicated models across three cancellation horizons, and why some members have no churn risk band at all.
Contents
- Churn Prediction is a separate feature
- The five risk bands
- New-joiner churn models (3M, 6M, 12M)
- Why a member may have no churn risk
Churn Prediction is a separate feature
Engage assesses every member daily based on historical cancellation patterns. What you see is a risk band, not a raw probability.
Churn Prediction is its own feature that must be booked and activated per studio, with its own configuration and permission. Once it is active, the risk bands are available to your Engage trigger rules automatically. If the churn triggers stay empty, check that Churn Prediction is active for the facility.
See also: MagicAI Engage overview and setup (Prerequisites) for the exact permissions and licensing (SETTINGS_CHURN_PREDICTION_CONFIGURATION, CUSTOMER_SHOW_CHURN_RISK).
The five risk bands
The risk bands, from lowest to highest risk, are:
| Band | Meaning |
|---|---|
| Very Low | Lowest assessed cancellation risk. |
| Low | Below-average cancellation risk. |
| Medium | Average cancellation risk. |
| High | Elevated cancellation risk. |
| Very High | Highest assessed cancellation risk. |
Bands are ordered, so a rule can compare against them. When an Engage rule says "Churn risk band ≥ High", it matches members in the High and Very High bands. The full ordering is: Very Low < Low < Medium < High < Very High.
New-joiner churn models (3M, 6M, 12M)
The general churn risk model needs visit history, so it cannot say anything meaningful about a member who just signed up. That early phase is exactly when outreach matters most, so new joiners get their own risk bands from dedicated new-joiner models, separate from the general churn risk model, at the earliest point a reliable prediction is possible: 42 days (6 weeks) after the contract starts, once the first weeks of behavior are available.
The new-joiner prediction looks at three cancellation time horizons: will this member cancel within 3, 6, or 12 months? That is why there are three models and three separate signals (3M, 6M, 12M), each with its own risk band using the same five bands. A member can be low risk on the 3-month horizon but high risk on the 12-month horizon. Before the day-42 scoring, all three fields are empty.
Each horizon has a matching default trigger rule (New Joiner Churn 3M, 6M, and 12M). See also: Creating and editing trigger rules.
Why a member may have no churn risk
A member has no band assigned, and is excluded from all churn-based triggers, if any of these apply:
- No marketing consent recorded (the Marketing consent category on the member profile)
- Not currently an active member (prospect or former member)
- Contract is cancelled, even if still within the notice period
- Under 18
- Not enough data (joined recently or insufficient visit history)
A member without a churn risk band is not broken. It usually just means the member is new (scored from day 42) or has not given marketing consent.