Define the decision
Write down who will use the output and what they will do differently. A model can perform well in isolation and still fail to support a useful decision. Agree what a workable outcome looks like before choosing an algorithm.
Examine the data
Check whether representative examples exist, whether labels are reliable and whether the data can be used for this purpose. Keep a record of gaps, ownership and access constraints.
Establish a baseline
Compare the proposed approach with a simple rule or the current process. A baseline helps you judge whether the additional complexity is worthwhile.
Decide what to test first
Choose a small experiment that could change your decision. Define success and stopping conditions in advance. A useful first outcome may be a clearer understanding of why development should wait.