Rogue Seasonality Detection in Supply Chains

Rogue seasonality or unintended cyclic variability in order and other supply chain variables is an endogenous disturbance generated by a company’s internal processes such as inventory and production control systems. The ability to automatically detect, diagnose and discriminate rogue seasonality from exogenous disturbances is of prime importance to decision makers. This paper compares the effectiveness of alternative time series techniques based on Fourier and discrete wavelet transforms, autocorrelation and cross correlation functions and autoregressive model in detecting rogue seasonality. Rogue seasonalities of various intensities were generated using different simulation designs and demand patterns to evaluate each of these techniques. An index for rogue seasonality, based on the clustering profile of the supply chain variables was defined and used in the evaluation. The Fourier transform technique was found to be the most effective for rogue seasonality detection, which was also subsequently validated using data from a steel supply network.