Predictive top-up models for Asian telecom operators
Predictive models for Asian telecom operators that steer prepaid top-up behaviour, using behavioural signals rather than blanket campaigns.
What we walked into.
Prepaid revenue lives or dies on when a subscriber recharges. Operators knew the aggregate, but not who was about to lapse and who simply recharges late.
Blanket reminder campaigns burn goodwill and margin at the same time.
Prepaid revenue is a timing problem. Knowing who is about to lapse is worth more than knowing the average.
The approach.
- Step 1
Modelled recharge behaviour per subscriber: interval patterns, balance decay, usage type and response to previous nudges.
- Step 2
Scored the base continuously, so the next best action is decided on today's behaviour instead of last month's segment.
- Step 3
Fed the scores into the operator's campaign channels, with holdout groups to keep the measurement honest.
Stack and disciplines.
What it delivered.
Recharge prompts pointed at the subscribers who actually respond to them, at the moment they respond.
Behavioural steering that lifts recharge activity without spraying the entire base.
Lessons that travel.
Blanket reminders burn goodwill and margin at the same time.
Scores must be fresh. A segment from last month describes a subscriber who no longer exists.
Holdout groups are the only thing that keeps a lift number honest.
Got something like this?
Tell us what you are building. We will be direct about whether Redwind is the right team for it.
Start a project