A logistics marketplace recommends spare carriers to shippers. The score for a shipper comes from which carriers similar shippers have booked, not from that shipper's fleet size, region, or contract type alone. Which method is the marketplace using?

Collaborative filtering recommends from patterns of behavior shared across users or across items. Here the score comes from which carriers similar shippers have booked. The information is in the booking relationships among shippers, not only in a description of the shipper being scored. That is the collaborative pattern, whether the neighborhood is similar shippers or, in the item form of the same idea, similar carriers. A content profile would score the shipper from attributes such as fleet size, region, and contract type. The stem says the score does not come from those attributes alone, so the content reading contradicts the method in use. A cold start is the absence of history for a new shipper or a new carrier, which leaves collaborative comparisons with nothing to read. The stem describes a score built from other shippers' bookings, which is history being used, not an empty history being bypassed. Counting carrier names as tokens and ignoring relationships is a bag-of-words style encoding of text. It does not use the co-booking pattern across shippers that the marketplace is using. Exam caveat: a production ranker can blend collaborative signals with attributes, but the signal this stem isolates is the shared booking pattern. Operational check: trace a recommended carrier back to similar shippers' bookings, and confirm the score still moves when fleet size, region, and contract type are held fixed.