Churn Prevention Intelligence · Wisdom notebook
Ask the follow-up.
Agentic conversational BI · from the score movement to the expiry window
Which subscribers moved more than 20 churn-score points in the last two weeks, and what are they worth?
RKFiltered the scored table for a score delta above 20 points over 14 days, then joined each subscriber's own ARPU and tenure to size the exposure by cohort.
About 41,000 subscribers, but the value is very unevenly spread — four cohorts hold most of it:
Revenue at risk per month, by cohort
18,000 subscribers — 44% of those who moved — carry less than a tenth of the revenue at risk.
Is the cause the same in each of those cohorts?
RKChecked our own signals against each cohort — care complaints by category, session quality, billing disputes and tariff changes — and kept only what moved with the score.
No. Three different causes, and only one of them is a retention problem:
New York is network experience, Chicago is a billing-dispute problem, Los Angeles is genuinely price. Same rising score, three owners.
Break the at-risk base by contract-expiry window — how much is inside 60 days?
RKJoined the scored base to commitment end-dates and handset-instalment schedules, bucketed by weeks to expiry, and sized the revenue sitting in each bucket.
$2.6M of the $4.1M reaches its commitment end inside 60 days — and the accelerating cohorts sit in the same buckets:
Revenue at risk by weeks to commitment expiry
Risk without an expiry date is not yet a port-out. Two-thirds of the exposure clears its commitment within eight weeks — and in the 5–8 week bucket the handset instalment ends at the same time, which is the week the number actually leaves.