10%Simulation
Proving an alert is worth sending
of non urgent alerts held back is enough to detect a 30% cut in arrears
- What was observed
- An early warning system raises alerts, and a team works them.
- What a generic approach says
- Arrears fell after the calls, so the calls worked.
- What the engine read
- Holding back a random 10% of non urgent alerts for a year detected a 30% cut in later arrears in every simulated run. Urgent alerts are never held back.
- What it meant for the decision
- The value of acting is measured against borrowers who were not worked, not assumed from what happened next.