Churn Risk

The Churn Risk tab scores every active, on-hold and pending-cancellation subscription from 0 to 100 and shows you the signals behind each score, so you can act on the subscribers most likely to leave before they do.

Churn scoring is not an AI feature. It is a deterministic weighted sum of signals your store already has, it runs entirely on your own server, and it makes no external requests — whether or not AI Insights is enabled.

Summary cards

CardWhat it means
High riskSubscriptions scoring 70 or above — the strongest combination of risk signals: failed payments, on-hold status, or an already-scheduled cancellation.
Medium riskSubscriptions scoring 40 to 69. Worth watching: usually one meaningful signal, such as a single failed renewal or a manual-renewal subscriber.
Low riskSubscriptions scoring below 40. No significant risk signals detected.
MRR at riskCombined monthly recurring revenue of every high-risk subscription — what you stand to lose if none of them recover.

The subscriber table

Each row is one subscription, with:

  • Subscription — the subscription number, linking to the subscription itself.
  • Customer — the subscriber’s name.
  • Product — what they subscribe to.
  • Status — Active, On hold, or Pending cancellation.
  • MRR — that subscription’s monthly recurring value.
  • Risk — the 0 to 100 score and its band.
  • Why — the specific signals that produced the score.

Use the Show filter above the table to narrow it to All, High, Medium or Low.

The signals and their weights

The weights are published deliberately: a score you can inspect is a score you can trust and argue with.

Signal shown in WhyPointsWhen it applies
Cancellation already scheduled95 (fixed)The subscription is pending cancellation. This is not a prediction — it has already happened.
Subscription on hold25The subscription status is on-hold, usually after a failed payment.
N failed renewal payments15 each, capped at 35A renewal payment failed within the last 90 days.
Payment overdue by N days20The scheduled payment date passed more than 3 days ago. The grace period stops a gateway retrying overnight from being flagged.
No successful renewal yet15The first renewal was due and never landed.
Still within its first billing cycle10Too early to have renewed yet — mild early-life risk.
Manual renewal, no saved payment method10The subscriber has to pay manually each cycle, which many never do.
Early-life risk keys off whether the subscriber has proven they will renew, not off raw age. One successful renewal clears both early-life signals — so a 400-day-old annual subscription is not treated as new just because two cycles of a yearly plan is 730 days.

Tuning the weights

The weights can be adjusted for your store with the zorem_subs_churn_signal_weights filter, which receives an array of signal code to weight. Scores are recalculated on the next run.

When scores are calculated

  • Nightly — every open subscription is rescored.
  • Shortly after a state change — when a subscription goes on hold, a payment fails, or a cancellation is scheduled, that subscription is rescored without waiting for the next night.
  • On demand — the Recalculate button at the top of the tab rescores everything immediately.

The first time you open the tab you will see No risk scores yet with a Calculate risk scores button. Click it rather than waiting for the nightly run.

Using the report

  • Failed payments and overdue subscriptions are the most recoverable — a dunning email or a request to update the card often saves them.
  • Manual-renewal subscribers are worth converting to automatic payment; that single change removes a standing 10-point risk.
  • MRR at risk is the figure to take to a revenue meeting. It is the concrete cost of not acting.