A new grey prediction model for the return material authorization process in the TFT-LCD industry

In the era of the short product life cycle, the return materials authorization (RMA) process plays an important role in satisfying customers because sellers have to prepare a certain amount of stock for returning products. Hence, determining how to evaluate the trade-off between over-stock and under-stock to balance cost-cutting and customer satisfaction is worthy of study. In this paper, we reveal a case taken from a leading TFT-LCD (thin film transistor liquid crystal display) maker in Taiwan, where the prepared stock is very likely to go from being hoarded to futility because the life cycle of TFT-LCD products is very short. To overcome the issues of the RMA process in the case company, obtaining the precise sizes of returning defectives in the near future is considered as one of the key tasks. General grey models (GMs) are widely applied to make predictions for this kind of short-term time series data; nevertheless, there still has the possibility of improving its predictions. This study thus proposes a new GM model, called the MTD-GM model, where a mega-trend-diffusion technique is employed to improve estimating the background values in the traditional GM. The experimental results show that the proposed MTD-GM grey model outperforms four other grey models with eight products in the case company data.