A consultant is preparing to enable AI-driven predictive opportunity scoring for a sales organization migrating to Dynamics 365 Sales. (Select TWO.) Which two conditions should the consultant confirm before turning the model on so that it produces reliable scores?
Select all correct answers, then click Submit.
Short Explanation
Think of this model as a student who can only learn from the case studies it has actually seen, so it needs plenty of examples of both outcomes, deals that closed and deals that fell apart, to figure out what separates the two. If the historical pile is mostly success stories, or barely has any history at all, the lessons it draws will be shaky. On top of that, the fields it studies from, things like when a deal is expected to close, what it is worth, and how it moved through stages, need to actually be filled in consistently across those old records, because a model cannot learn patterns from blank fields any more than a student can learn from missing pages. Worrying about whether every record owner holds a particular license, or stripping out custom fields before flipping the switch, are not real prerequisites here; they just are not how the readiness of this kind of model actually gets determined. Data volume, balance, and completeness are the two things worth checking first.
Full Explanation
The correct answers are B and C. Predictive opportunity scoring is a machine learning model trained on historical opportunity data, so it needs a sufficient volume of past opportunities with both won and lost outcomes to learn what distinguishes a deal that closes from one that does not; a dataset skewed toward only wins, or too small, produces an unreliable model. It also depends on core fields like estimated close date, estimated revenue, and stage history being populated consistently across those historical records, since gaps or inconsistent entry starve the model of the signals it needs to score new opportunities meaningfully. Confirming both conditions before enabling the model is standard pre-flight due diligence during a migration. Option D is incorrect because the model scores opportunity records based on their data, not the licensing tier of the record owner; there is no such licensing gate on which records get scored. Option A is incorrect because predictive scoring is not restricted to out-of-box fields only, and disabling custom fields is unnecessary and would only remove potentially useful signals rather than help the model.