A predictive model is not a decision system
A risk model gives you a surface of probabilities. Operations needs a small set of deployable actions.
The Problem
In a Nairobi ambulance deployment competition, the model predicted the probability of a crash for each road segment × 3-hour window.
But that was not the actual output the problem needed. The real decision was: Where should 6 ambulances be positioned right now?
That is a different object entirely. A risk model gives you a surface of probabilities. Operations needs a small set of deployable actions.
The Translation Layer
The important part was the translation layer between prediction and action:
- Predict crash risk for every segment in the city
- Select the highest-risk segments for the upcoming window
- Cluster those predicted hotspots into 6 geographic groups
- Place ambulances at the centroids
The model did not decide where ambulances go — it produced a risk map. The deployment layer converted that map into an operational plan.
The Pattern
Fraud scores
vs. investigation capacity
Hospital risk scores
vs. nurse allocation
Demand forecasts
vs. inventory placement
Outage probability
vs. field crew dispatch
Key Takeaway
Design insight: In real-world ML, prediction quality and decision quality are related, but they are not the same thing. The model predicts. The system decides.