One prepaid cohort is churning at nearly twice its own rate, and the network in its own cells is why
Churn on the prepaid book in the South reached today. Every other region is still running at . The pre-churn score on the same cohort has climbed to . Most of that climb is one input: network experience reached , because latency in the cells this cohort uses got worse through the same three weeks. Subscribers called about it, and the complaint input reached . The usage and billing input eased, to . And port-outs has not moved at all yet, which is the only reason any of this is still actionable.
So what? Most of what the score is reading is the network in these cells, and a discount does not move a network reading. Nobody has actually gone yet. Once the number ports, the handset and the household go with it.
3.1 vs 1.8% three weeks ago
nearly twice its own rate
58 vs 42
most of it one input
31 vs 18
thirteen of the score's sixteen
19 vs 12
they called rather than left quietly
1.8 vs 1.8%
one cohort, not a market
on its own level
decided, not yet gone
Churn Rate by region, against each region's own level
One cohort has moved. The rest of the base has not.
The line on each bar is that row’s own expected level. The scale runs from 0 to 3.5.
The prepaid book in the South reached , where every other region sits at . One cohort has moved and the rest of the base has not, which is what makes this a cohort to work rather than a market to react to.
Churn Rate and Pre-Churn Score, three weeks
The score moved first. The churn followed.
Churn on the cohort started the three weeks at and finished them at , climbing in each one. The pre-churn score reached over the same stretch and led it, which is what the score is for.
The score's three inputs, and the two readings that held
Each input against what it normally contributes to the score.
| Reading | Across the three weeks | What it says |
|---|---|---|
| Pre-Churn Score · network experience | 31 against its own 18 | latency in the cohort's own cells got worse |
| Pre-Churn Score · complaints | 19 against its own 12 | subscribers called about it |
| Pre-Churn Score · usage and billing | 8 against its own 12 | not a bill and not a usage drop |
| Tenure Mix | held on its own level | not a change in who is in the cohort |
| Port-Out Rate | held on its own level | nobody has actually gone yet |
| What is left | the network in those cells | a network fix, not a save offer |
The score's network-experience input reached , most of its climb, because latency in the cells this cohort uses got worse through the same three weeks. The complaint input reached , so subscribers called about it rather than leaving quietly. The usage and billing input eased to and tenure mix held on its own level, which takes price and cohort drift off the table.
3.1 vs 1.8%
the rate still running
31 vs 18
what those cells are costing
on its own level
the window, while it lasts
One cohort, and the rest of the base is untouched
Churn on the prepaid book in the South reached while every other region sits at .
Why it matters: It is a network fix in one region rather than a retention programme for the base.
The score's largest input is the network
Network experience reached and the complaint input , while the usage and billing input eased to .
Why it matters: Latency in the cohort's own cells is what those two inputs read. A cohort leaving over coverage is not held with a save offer, and the fix sits with another team.
Nobody has actually left yet
The pre-churn score reached while port-outs has not moved at all yet.
Why it matters: The score names the subscribers one by one and the port-out rate says they are all still reachable. Both are true only today.
Which subscribers in the cohort carry the highest pre-churn score?
Turns a cohort into a call list ordered by risk.
Which cells in the South are the at-risk subscribers actually on?
Hands the network book a list of cells rather than a region.
What did port-outs do the last time this cohort's score climbed this far?
Says how long the window actually is.
Is any other region's network-experience input starting to climb?
Uses the leading reading as a leading reading.
What tenure do the at-risk subscribers have?
Decides whether a save offer is worth what it costs.