Despite increasing competitive pressure and shrinking profit margins, air cargo carriers currently do not widely make use of dynamic pricing: price quotes in air cargo are often negotiated through human interaction. At Lufthansa Cargo, a machine learning model is currently in development to automate spot quote pricing. It learns from shipment and market data to determine booking probabilities, but customer-specific information is currently not taken into account. Therefore, in this thesis, it is examined which customer-specific features can be modeled, whether clusters of customers can be revealed by customer data, and if customer-specific features improve the ability to predict booking probabilities. Following a feature engineering process, a k-means clustering is performed to find groups of customers with different willingness to pay. Then, the machine learning model is extended by these new customer features. It is evaluated through the use of metrics, calibration and conversion rate curves: do the model’s predictions improve? The results show that while clusters with different willingness to pay exist, meaningful differences are limited. On the other hand, it is shown that select customer features significantly improve model performance. Especially the model recall – its ability to predict actual bookings correctly – has improved. Further research potential is seen in (i) the more refined modeling of customer willingness to pay and in (ii) the evaluation of customer behavior, by the use of booking platform metadata.