A consultant is enabling predictive opportunity scoring for a manufacturing holding company whose Dynamics 365 Sales instance serves two business units: a direct-sales industrial unit with six years of consistently tracked opportunity history, and a distributor-channel unit that was onboarded four months ago with only a handful of closed opportunities. Leadership wants one dashboard showing scores for every open opportunity across both units. What should the consultant do?
Select an answer to reveal the explanation.
Short Explanation
Think of a scoring model like a forecaster who needs a long run of past seasons before its predictions mean anything. One business unit here has years of that history, so its forecasts can be trusted. The other unit barely has any season behind it yet, so whatever number came out would just be noise dressed up to look precise. The tempting shortcuts don't actually fix that: blending the thin unit into the big one waters down the good model with unreliable signal, and inventing history for the thin unit by copying someone else's outcomes just teaches the model a fiction. Swapping in a different AI feature altogether sidesteps the real question instead of answering it. The steadier move is to let the unit with real history get real scores now, and hold off on the newer unit until it has built up enough of its own closed deals to learn from.
Full Explanation
The correct answer is B. Predictive opportunity scoring learns from a segment's own closed won and lost opportunities, and the channel unit simply has not generated enough of that history yet; scoping the model to the direct-sales unit, which has six years of consistent data, keeps scores meaningful while the channel unit matures. Option A is incorrect because pooling the channel unit's thin, recent data into an organization-wide model would dilute the training set and produce scores that look confident but do not reflect either unit's actual patterns. Option C is incorrect because copying another unit's outcomes into the channel unit's records fabricates history rather than reflecting what actually happened there, corrupting the model and misleading anyone who reviews that unit's pipeline. Option D is incorrect because it substitutes a different capability for the one leadership asked about; Copilot's opportunity summary condenses activity on a single record and was never meant to replace a scoring model across a whole unit's pipeline, so it does not solve the underlying data-volume problem.