Determining the dataset

To do so, we worked with a detailed transactional dataset tracking 70 United States shippers’ dry van truckload shipments (moving greater than 250 miles) from 2015 to 2019. The data included shipment contract price, pickup and drop-off locations, lead time, carrier attributes and whether the shipment was accepted by the contracted carrier (and if not, whether a backup carrier accepted the shipment or if it went on the spot market). To better inform carrier acceptance decisions, we built a highly accurate predictive model that determines the probability that an individual load will be accepted by a contracted carrier given a set of attributes known based on our previous research. These attributes included whether the contracted (and backup) carrier is an asset-based provider or a broker, lane demand patterns (such as consistency and frequency of tendered volume to the contracted carrier), whether the shipment is considered surge volume above and beyond the expected weekly volume on that lane and – importantly – contract price relative to the lane-specific market price at that time. This final element allows us to measure carrier acceptance rates (and thus, expected price a shipper pays due to rejections) under our two pricing scenarios of interest: traditional fixed-price contracts and dynamic-priced contracts (as illustrated in Figure 2).