You are considering a new predictive feature for a connected heavy-equipment platform that would use machine learning to identify machines likely to need service soon, so dealers and fleet managers can act earlier. The feature would only be worth building if the model is accurate enough to change maintenance decisions and create measurable product value.
How would you decide whether this feature is worth investing in, and how would you approach building and evaluating the model so you can make that decision with confidence?