Cost to serve models in the industrial gas industry
In the industrial gas industry, companies mainly use truck trailers to deliver liquid product (NI, OX, AR, etc.) to the customers. The transportation cost is a substantial component of the total cost to serve the customer. In the competitive environment, the real cost to serve customers is a critical indicator to the sales department. A reasonable and fair cost allocation method can lead to a more reliable pricing policy not only for winning the market but also making reasonable profits. In this dissertation, we consider the problem of estimating the cost to serve individual customers in the industrial gas industry: a key issue is an appropriate allocation of the transportation cost among multiple customers on a route. Three sub problems are related to this problem. First, in a typical multiple stop delivery route, the customers are filled with different volumes of the product; the tank size of each customer is different; and the distances from customers to the plant and to each other are different. The task is to allocate the transportation cost to each customer fairly. A moat model is formulated to solve this problem; the moat model takes cooperation between customers into consideration and shares some cost based on measure of satisfaction with the fill volumes. We developed a complete strategy for application which can handle all kinds of scenarios. Second, for the new customer, there is no route information available; we need to generate the route candidates for the new customer. A depth first search method is developed to generate the routes. In growing and searching the tree, we also applied volume and distance fathoming rules to abandon growing impossible routes to increase the search speed when generating all possible candidate routes. Third, for a new customer, there is no history of real operation information such as the frequencies of real routes used and fill volumes available. Furthermore, a "perfect" route in the model could not be accepted in real practice because the customers on the route cannot accept product on the same day. Our task is to forecast the frequency of routes and fill volumes which will be used to approximate the future reality as well as possible. A MILP (mixed integer linear programming) model is developed to forecast the route frequency and combined with the moat model to obtain the cost to serve the new customer.