A consultant is preparing to enable predictive opportunity scoring for a division whose Dynamics 365 Sales instance has been live for two years. (Select TWO.) Which two actions should the consultant take before turning the model on?
Select all correct answers, then click Submit.
Short Explanation
Getting ready to switch on the scoring model is a bit like preparing a new analyst before they start making calls: you need to make sure they actually have the case files to study, and you need to make sure they are cleared to sit at the desk and do the job in the first place. That means checking that there is enough clean historical data covering both wins and losses, and checking that the right licensing and access are actually turned on for this environment. Typing in your own guess at a win percentage for every open deal does not teach the model anything, since it learns from what actually happened in the past, not from someone's manual estimate. And wiping out every lost deal would strip away half of what the model needs to learn, because understanding what a loss looks like is just as important as understanding what a win looks like. The two real prerequisites are solid historical data and the access to actually run the feature.
Full Explanation
The correct answers are A and B. Predictive opportunity scoring needs a sufficient body of historical won and lost opportunities with clean, consistently populated fields to learn real patterns, so verifying data volume and quality is a necessary pre-check described in option A. It also depends on the appropriate licensing and security roles being provisioned in the environment so the model can actually be trained and executed, which is the check described in option B; without that access in place, the feature cannot be turned on regardless of how good the data looks. Option C is incorrect because manually assigning probability numbers to open opportunities does not create a training baseline for the model; the model trains from historical closed outcomes, not from estimates a person types into open records. Option D is incorrect because deleting every lost opportunity would remove exactly the negative examples the model needs to learn what a losing deal looks like, leaving it unable to distinguish winning patterns from losing ones and severely degrading its accuracy rather than improving it. The two genuinely necessary preparatory steps are validating the historical data and confirming the licensing and access are configured.