(Select TWO.) A consultant is preparing to turn on predictive opportunity scoring for a client's Dynamics 365 Sales environment. Before enabling the feature, the consultant reviews the environment's data. Which two conditions should the consultant confirm are true so the resulting scores are reliable?
Select all correct answers, then click Submit.
Short Explanation
Getting useful scores out of a predictive model is a lot like teaching someone to judge deals by showing them a big pile of past examples, both the ones that closed and the ones that fell through, and making sure every example in that pile has the same key details filled in consistently. Skimp on either the number of past examples or the consistency of those details, and whatever pattern gets learned will be shaky. Whether every current deal has a contact person assigned, or whether reps are still typing in their own gut-feel percentage alongside the AI's number, doesn't actually change how well the model learned its patterns in the first place. Those are separate housekeeping questions, not the foundation the scoring itself is built on.
Full Explanation
The correct answers are C and D. Predictive opportunity scoring is a learned model, so it needs a meaningful volume of past won and lost opportunities to recognize patterns, and it needs the fields it actually uses, such as estimated value, close date, and stage progression, to be populated consistently across records; without both, the model either has too little to learn from or is learning from gaps and inconsistencies that produce unreliable scores. Option A is incorrect because a primary contact assignment, while good CRM hygiene, is not one of the data conditions the scoring model depends on, so its completeness does not affect score reliability. Option B is incorrect because manual probability-of-close entry and the AI-generated score are separate fields that can coexist without conflict; there is no requirement to disable manual entry, and doing so would remove information reps may still want to record rather than improve the model's accuracy.