Before turning on predictive opportunity scoring for a client, a consultant reviews the Dynamics 365 Sales data to judge whether the model will produce trustworthy results. (Select TWO.) Which two conditions should the consultant confirm are true?
Select all correct answers, then click Submit.
Short Explanation
Before you trust a model that is supposed to tell you which deals look like winners, you have to ask what it actually learned from. It learned from history, so two things matter: did the closed deals get labeled honestly as won or lost, and were there enough of them to show a real pattern rather than a fluke. Get either of those wrong and the score coming out the other end is just noise dressed up as insight. Whether every open deal right now happens to have a contact filled in is a different question about relationship tracking, not about whether the training history was solid. And whether reps sat through a certification course says nothing about the quality of the historical outcome data either, it is a people-readiness question, not a data-readiness one. The two that actually determine trustworthiness are the accuracy of past outcome labeling and having enough of that history to learn from.
Full Explanation
The correct answers are A and D. Predictive opportunity scoring learns by finding patterns that distinguish opportunities that were actually won from ones that were actually lost, so it depends on closed records having an accurate Status Reason: without that, the model has no reliable outcome label to learn from. It also depends on there being enough historical closed opportunities to detect a real pattern rather than noise from a handful of records, since a model trained on too few outcomes will not generalize to new, open opportunities. Both conditions speak directly to whether the training data is sufficient and trustworthy. Option B is incorrect because whether an open pipeline opportunity has an assigned primary contact affects relationship tracking, not whether historical outcome data exists to train the model; scoring can be enabled and calibrated correctly even if some open records lack a listed contact. Option C is incorrect because reps completing a certification course has no bearing on the historical data quality or volume that predictive scoring depends on; it may be good practice for adoption, but it is not a prerequisite the model itself requires to produce trustworthy scores.