A distributor completes an acquisition of a smaller competitor and merges its customer base into the parent company's Dynamics 365 Sales environment. The acquired unit has only three months of deal history in the system. Within weeks, the AI-driven opportunity scoring model begins assigning most of the acquired unit's opportunities a score near the middle of the range, regardless of deal specifics, while scores for the parent company's long-established pipeline remain sharply differentiated. A consultant is asked why the acquired unit's scores are not useful for prioritization. What should the consultant recommend?
Select an answer to reveal the explanation.
Short Explanation
Think of opportunity scoring as a student who learns from examples. Give it thousands of past deals with known outcomes and it starts recognizing the patterns that separate winners from losers. Give it only a few months of history, like the newly acquired unit here, and it simply has not seen enough examples yet, so it plays it safe and scores everything close to average. That is not a bug, it is the model honestly telling you it does not have enough evidence. The trap answers try to force a fix before there is enough data: retraining on the same thin slice of history just teaches the model the same limited lessons again, turning off scoring everywhere throws away a tool that is working fine for the rest of the business, and manually forcing every record to the same number just hides the problem instead of solving it. The practical move is patience paired with a fallback: let sellers on the new unit qualify deals the old-fashioned way until the pipeline builds up enough history for the model to speak with confidence.
Full Explanation
The correct answer is B. AI-driven opportunity scoring learns to differentiate strong and weak deals by finding patterns in a large set of historical won and lost opportunities; with only three months of history, the acquired unit has not yet given the model enough examples to learn what separates a deal likely to close from one unlikely to close, so scores cluster near the middle rather than spreading out. The consultant should explain this data-volume limitation and recommend that sellers on the acquired unit fall back on manual qualification criteria until enough deal history accumulates for the model to produce meaningful differentiation. Option A is incorrect because retraining on only three months of data does not solve a cold-start problem; the model still lacks the volume of outcomes needed to learn reliable patterns, so retraining now would not improve the acquired unit's scores. Option C is incorrect because the parent company's long-established pipeline is already producing useful, differentiated scores, and disabling scoring organization-wide would remove value from users where the model is working. Option D is incorrect because assigning every acquired-unit record the same fixed score removes the value of scoring entirely and gives sellers no prioritization signal at all, which is worse than acknowledging the limitation and using manual criteria in the interim.