* Corresponding author. Tel.: #65 874 3182; fax: #65 779 2621; e-mail: fbasumcc@nus.edu.sg. Int. J. Production Economics 58 (1999) 303— 318 Analysing interaction effects on MRP implementation using ACE Chee-Chuong Sum*, Ser-Aik Quek, Hoon-Eng Lim Department of Decision Sciences, Faculty of Business Administration, National University of Singapore, Singapore Received 24 November 1997; accepted 6 August 1998 Abstract This research represents the first study to examine the significance of interactions among individual determinant variables in affecting MRP implementation outcome. A powerful statistical technique, alternating conditional expecta- tions (ACE), was used to maximise the model fit and to uncover underlying non-linear relationships between implemen- tation success and its determinant variables. The final ACE model indicates that interaction effects are significant in affecting MRP implementation outcome. The analytic transformations from the ACE model presented new insights and information into the MRP implementation process. The managerial implications arising for the findings are also discussed.  1999 Elsevier Science B.V. All rights reserved. Keywords: MRP implementation; Alternating conditional expectations (ACE); Interaction effects; Non-linear regression models 1. Introduction When properly implemented, MRP systems can offer enormous benefits in areas such as costs, cus- tomer service, production scheduling, inventory, and functional co-ordination [1—5]. Successful MRP users have reported as much as 15% gains in manufacturing productivity, 50% reduction in overtime, 33% reduction in inventory investments, and 80% reduction in inventory shortages [6]. However, cases of successful MRP implementa- tion are not common [3,7]. Anderson et al. [8] estimated that less than 1 in 4 implementations is successful, while other studies reported that more than 50% of all implementation were disappoint- ments [9,10]. Unsuccessful MRP implementation not only deprive companies of potentially huge benefits, but may also result in financial losses and disruptions in operations [1,11,12]. The current literature on the theoretical model- ling of the MRP implementation process has prim- arily focused on examining the determinant variables separately without regard for interactions among the determinant variables. For example, variables such as top management support, pro- duction support, and data accuracy were examined separately as main effects in the modelling process. There exists to date no MRP study that assesses the significance of interaction effects among variables within the theoretical frameworks of 0925-5273/99/$ — see front matter  1999 Elsevier Science B.V. All rights reserved PII: S 0 9 2 5 - 5 2 7 3 ( 9 8 ) 0 0 2 0 5 - 9