Rodrigo Cambiaghi Azevedo, Sophie D’Amours, Mikael Ronnqvist

Order promising is one of the most important activities for all manufacturing organizations. It is the moment of truth when customer service level and value chain profitability are realized. ERP systems have introduced the ATP (available-to-promise) feature in order to support promise makers with accurate supply information which would guarantee a correct delivery date promise to customers. When APS systems were later introduced, the order promising activity gained additional support such as the ability to re-plan manufacturing settings, if necessary, every time an order comes in (called capable-to-promise), the use of rules to search for alternative products, supply location or a combination of both, and online check of the order quantity against previous allocated quotas or forecasts.

However, the recent success of yield management for service industries such as airlines, hotels and car rentals, has opened up a field for profitability maximization when promising orders to customers.

Essentially, its success is related to the capacity the companies developed to identify and segment their customers according to their individual willingness to pay. Once customers are segmented according to different profitability levels, companies are able to reserve products for more profitable customers in preference to less profitable ones. To accomplish this, the concept of allocated available-to-promise (aATP) was developed.

In his pioneering work, Meyr (2008) uses data from light industry to introduce and demonstrate the potential benefits of the aATP and real-time order promising configuration for make-to-stock manufacturing environments. He also discusses the importance of determining a reasonable number of customer segments in order to either avoid loss of profits with too few segments or increase forecasting complexity with too many segments.

In this paper, we aim to further develop this concept. First of all, we apply the aATP and real-time order promising proposition to a Canadian softwood lumber manufacturer. This application allows us to analyze the concept from a network perspective (several storage locations and different geographical customer segments) as well as in a commodity-type market. In addition, we further test the aATP concept by adding two analyses still not available in the literature: first, we analyze the impact on the profitability performance when individual orders request more than one product. A drop in the profitability performance was found once the inventory is allocated based on independent product-segment forecasts. Second, we test the performance of the aATP concept after introducing different levels of forecast errors. Meyr (2008) does not consider forecast errors in his analyses. His justification for this is that allocation planning activities are performed periodically (e.g. weekly) using rolling horizons when the variabilities are corrected. By adding certain level of forecast errors to the allocation planning we found that the company’s profitability might deteriorate even between two consecutive allocation planning activities.

The analyses presented in this article were performed using four comparative order-promising optimization models developed in the AMPL mathematical language. A total of 440 scenarios were evaluated based on a database of 81 different orders.