A consultant configures the similar-opportunities recommendation for a client that sells industrial equipment across several unrelated verticals, food processing, mining, and pharmaceuticals, through one Dynamics 365 Sales instance. Reps in the mining division report that the similar-opportunities suggestions on their deals mostly reference unrelated food-processing deals with little practical relevance. What should the consultant adjust?
Select an answer to reveal the explanation.
Short Explanation
When a recommendation feature keeps surfacing deals that have nothing to do with what a rep is actually working, the fix usually is not to give up on the feature, it is to look at what it is matching on. If it is comparing deals without regard to industry or product line, reps in totally different markets will keep seeing each other's opportunities, which is exactly the complaint here. Tightening those matching criteria so mining deals compare against other mining deals solves it directly. Turning the whole thing off throws away something that could be genuinely useful once scoped correctly, and splitting the client into separate environments per vertical is a massive structural overreaction to what is really just a configuration tweak. And clicking not similar on individual bad suggestions one at a time is not how this feature learns; it does not retrain itself from that kind of manual feedback.
Full Explanation
The correct answer is D. The similar-opportunities recommendation compares deals using configurable matching criteria, and when reps in one vertical keep seeing suggestions from a completely unrelated vertical, the fix is to review and adjust which fields, such as industry or product line, the feature uses to judge similarity so results are scoped to genuinely comparable deals. Option A is incorrect because it assumes the feature has no scoping controls and must always blend results across verticals, which is not accurate; the matching criteria can be adjusted rather than requiring the feature to be turned off entirely. Option B is incorrect because splitting the client into separate environments per vertical is a drastic structural change unnecessary for what is fundamentally a configuration adjustment, and it further wrongly claims the similarity fields cannot be reconfigured, contradicting the actual fix. Option C is incorrect because the feature does not learn from individual reps manually flagging suggestions as not similar in the way described; that is not how its matching mechanism operates, so this approach would not change future recommendations.