1. Accuracy: What is the cost of being wrong?
2. Explainability: Will you be able to explain the decisions your model makes?
3. Reproducibility: Does your model need to generate the same answer every time?
4. Confidentiality: Will you carry the risk if private information becomes public?
5. Data quality: Is the data you will use correct and complete? (What is the source of truth?)
6. Generalizability: Does the training (data) reflect the operating (data) conditions?
Adapted from George Westerman et al, Find the AI Approach that Fits the Problem You're Trying to Solve.