A consultant is configuring predictive lead scoring for an insurance brokerage that sells both auto and commercial policies through the same Dynamics 365 Sales environment. Auto leads close in days with high volume and low deal value, while commercial leads close over months with low volume and high deal value. After enabling scoring, the auto team reports the grades feel accurate, but the commercial team says nearly every lead lands in the same middling band regardless of outcome. What should the consultant do to fix the commercial team's scoring without disrupting the auto team?
Select an answer to reveal the explanation.
Short Explanation
When one sales motion moves fast with lots of small deals and another moves slow with a handful of big ones, cramming both into a single scoring model is like asking one thermostat to keep a sauna and a walk-in freezer at a comfortable temperature. The model ends up splitting the difference, and the slower, higher-value deals all get shoved into a lukewarm middle grade that does not actually tell the team anything useful. The fix is not to starve the model of data or throw the whole system out because one segment is not working, and it is not to just turn up one signal's influence while still forcing both business lines through the same brain. The real fix is giving the slower-cycle business its own dedicated model that only ever sees its own history, so it can learn what actually separates a strong commercial prospect from a weak one on its own terms. That leaves the fast-moving segment's already-working setup completely alone while giving the other team scores they can trust.
Full Explanation
The correct answer is A. Auto and commercial leads convert on fundamentally different patterns and timelines, so a single shared model trained across both segments learns an averaged pattern that fits neither well, which is why commercial leads cluster in a flat middle band despite real differences in quality. Scoping a second model to only the commercial leads lets it learn the signals and timing that actually predict commercial conversion, while leaving the existing auto model untouched. Option B is incorrect because lowering the data threshold makes an underpowered model even less reliable rather than addressing the mismatch between segments. Option C is incorrect because disabling scoring organization-wide sacrifices the auto team's working model to fix a problem that only affects commercial leads. Option D is incorrect because reweighting a signal within the same shared model still trains on both segments together, so it will not separate the distinct behavioral patterns driving auto versus commercial conversion.