Churn prediction.

Every active subscriber gets a 0–100 risk score, recalculated daily from behavioral signals. Reach out to the high-risk subscribers before they hit cancel — recover revenue you'd otherwise lose.

Last updated · April 27, 2026

How the score is calculated

Daima reads behavioral signals from your subscribers and combines them into a single score from 0 to 100. Higher score = higher risk of cancelling. The signals include:

• Portal visits. A spike in customer-portal activity (especially viewing the cancel flow) is a strong precursor to cancellation.
• Skip and pause activity. Subscribers who skip 2+ deliveries in a row are statistically more likely to cancel within 60 days.
• Plan downgrades. Switching to a less frequent delivery cadence (e.g. monthly → bi-monthly) signals reduced engagement.
• Missed renewals. Failed payments where the retry hasn't been resolved.
• Order history shape. Long-tenured subscribers are weighted lower (they've already shown commitment); newer subscribers get higher base risk.

Risk tiers

The Churn Prediction admin page groups subscribers into three tiers based on their score:

• High Risk — score 70 or above. Subscribers who likely cancel within the next 30–60 days unless intervened with.
• Medium Risk — score 40 to 69. Watch list. Worth a check-in but not urgent.
• Low Risk — score below 40. Healthy subscribers. No outreach needed.

The page defaults to showing the High Risk list — that's the actionable cohort. Switch the filter to see all subscribers, or just one tier.

Daily recalculation

Scores recalculate automatically once per day for every active subscriber. New signals (a portal visit overnight, a skip, a missed renewal) factor into the next day's score.

You can trigger a manual recalculation by clicking Refresh scores on the Churn Prediction page. Useful if you've just made a change you want reflected immediately (a price change, a product change, a support outreach you want to see weighted).

Acting on the high-risk list

The high-risk list shows each subscriber's name, score, last meaningful activity (e.g., "Skipped last 2 deliveries"), and a link to their contract detail. From the contract page you can:

• Email the subscriber directly (with their order history pre-populated)
• Apply a one-time discount on their next renewal as a retention offer
• Pause their subscription with a note instead of letting them cancel outright
• Mark the subscriber as "outreach completed" so they drop out of your active intervention queue

Designing a churn-rescue workflow

• Don't ignore Medium Risk. The High Risk list is small but late. The Medium Risk list is bigger and earlier — that's where retention dollars work hardest.
• Personalize the outreach. A generic "we miss you" email converts ~2x worse than one that references the subscriber's actual product or last order.
• Test discount levels. Some merchants find that a 25% one-time renewal discount saves 30% of high-risk subs. Some find that any discount feels desperate. A/B test it on retention emails too.
• Pause beats cancel. If a subscriber hits the cancel flow and you have an active retention email, offer them a 60-day pause instead. They keep the subscription open, you keep them in the funnel.

Honest limitations

Churn prediction is a probabilistic model — it doesn't know why a subscriber is at risk, just that they look statistically similar to subscribers who cancelled in the past. False positives happen (a subscriber who viewed cancel-flow but actually wanted to change billing date will look like a churn risk). False negatives happen too (subscribers who quietly let their card expire without any portal activity).

Use the score to prioritize your outreach, not to make automated decisions. The merchant-side workflow (you, looking at a list, deciding who to message) is what makes the prediction valuable. The score is a sort key, not a verdict.