Intermittent Demand Forecasting. John E. Boylan

and Cardiff

      January 2021

       John E. Boylan

       Aris A. Syntetos

Glossary

ADIDAaggregate–disaggregate intermittent demand approachAICAkaike information criterionARautoregressiveARIMAautoregressive integrated moving averageARMAautoregressive moving averageAPEabsolute percentage errorBObackorderBoMbill of materialsBSBrier scoreCDFcumulative distribution functionCFEcumulative forecast errorCSLcycle service level (all replenishment cycles)

cycle service level (replenishment cycles with some demand)CVcoefficient of variationEDFempirical distribution functionERPenterprise resource planningFMECAfailure mode, effects, and criticality analysisFRfill rateFSSforecast support systemFVAforecast value addedHEShyperbolic exponential smoothingINARinteger autoregressiveINARMAinteger autoregressive moving averageINMAinteger moving averageIPinventory positionKSKolmogorov–Smirnov (test)LTDlead‐time demandMAmoving averageMADmean absolute deviationMAEmean absolute errorMAPEmean absolute percentage errorMAPEFFmean absolute percentage error from forecastMASEmean absolute scaled errorMEmean errorMMSEminimum mean square errorMPEmean percentage errorMPSmaster production scheduleMROmaintenance, repair, and operationsMRPmaterial requirements planningMSEmean square errorMSOEmultiple source of errorMTOmake to orderMTSmake to stockNBDnegative binomial distributionNNneural networkNOBnon‐overlapping blocksOBoverlapping blocksOUTorder up toPISperiods in stockPITprobability integral transformRMSEroot mean square errorrPITrandomised probability integral transformS&OPsales and operations planningSBASyntetos–Boylan Approximation (method)SBCSyntetos–Boylan–Croston (classification)SCMsupply chain managementSESsingle (or simple) exponential smoothingSKUstock keeping unitSLAservice level agreementSMAsimple moving averagesMAPEsymmetric mean absolute percentage errorsMSEscaled mean square errorSOHstock on handSOOstock on orderSSOEsingle source of errorTSBTeunter–Syntetos–Babai (method)VZViswanathan–Zhou (method)WMHWright Modified Holt (method)WSSWillemain–Smart–Schwarz (method)

About the Companion Website

      This book is accompanied by a companion website.

      www.wiley.com/go/boylansyntetos/intermittentdemandforecasting

      This website includes:

       Datasets (with accompanying information)

       Links to R packages

1 Economic and Environmental Context

      1.1 Introduction

      Demand forecasting is the basis for most planning and control activities in any organisation. Unless a forecast of future demand is available, organisations cannot commit to staffing levels, production schedules, inventory replenishment orders, or transportation arrangements. It is demand forecasting that sets the entire supply chain in motion.

      Demand will typically be accumulated in some pre‐defined ‘time buckets’ (periods), such as a day, a week, or a month. The determination of the length of the time period that constitutes a time bucket is a very important decision. It is a choice that should relate to the nature of the industry and the volume of the demand itself but it may also be dictated by the IT infrastructure or software solutions in place. Regardless of the length of the time buckets, demand records eventually form a time series, which is a sequence of successive demand observations over time periods of equal length.

      On many occasions, demand may be observed in every time period, resulting in what is sometimes referred to as ‘non‐intermittent demand’. Alternatively, demand may appear sporadically, with no demand at all in some periods, leading to an intermittent appearance of demand occurrences. Should that be the case, contribution to revenues is naturally lower than that of faster‐moving demand items. Intermittent demand items do not attract much marketing attention, as they will rarely be the focus of a promotion, for example. However, they have significant cost implications for a simple reason: there are often many of them!

      Service or spare parts are very frequently characterised by intermittent demand patterns. These items are essentially components or (sub‐) assemblies contributing to the build‐up of a final product. However, they face ‘independent demand’, which is demand generated directly from customers, rather than production requirements for a particular number of units of the final product. In the after‐sales environment (or ‘aftermarket’), we deal exclusively with ‘independent demand’ items. Service parts facing intermittent demand may represent a large proportion of an organisation's inventory investment. In some industries, this proportion may be as high as 60% or 70% (Syntetos 2011). The management of these items is a very important task which, when supported by intelligent inventory control mechanisms, may yield dramatic cost reductions.

Industries that rely heavily on after‐sales support, including the automotive, IT, and electronics sectors, are dominated by intermittent demand items. The contributions of the after‐sales services to the total revenues of organisations in these industries have been reported to be as high as 60% (Johnston et al. 2003). This signifies an opportunity not only to reduce costs but also to increase revenues through a careful balancing of keeping enough in stock to satisfy customers but not so much as to unnecessarily increase inventory investments. There are tremendous economic benefits that may be realised through the reappraisal of managing intermittent demand items.

      There are also significant environmental benefits to be realised by such a reappraisal. Because of their inherent slow movement, intermittent demand items are at the greatest risk of obsolescence. The problem is exacerbated by the greatly reduced product life cycles in modern industry. This affects the planning process for all intermittent demand items (both final products and spare parts used to sustain the operation of final products). Better forecasting and inventory decisions may reduce overall scrap and waste. Furthermore, the sustained provision of spare parts may also reduce premature replacement of the original equipment.

      The area of intermittent demand forecasting has been neglected by researchers and practitioners for too long. From a business perspective, this may be explained in terms of the lack of focus on intermittent demand items by the marketing function of organisations. However, the tough economic conditions experienced from around 2010 onwards have resulted in a switch of emphasis from revenue maximisation to cost minimisation. This switch repositions intermittent demand items as the focus of attention in many companies, as part of the drive to dramatically cut down costs and remain competitive. In addition, the more recent emergence of the after‐sales business as a major determinant of companies' success has also led to the recognition of intermittent demand forecasting as an area of exceptional importance.

      Following a seminal contribution in this area by John Croston in 1972, intermittent demand forecasting received very little attention by researchers over the next 20 years. This was in contrast to the extensive research conducted on forecasting faster‐moving demand items. Research activity grew rapidly from the mid‐1990s onwards, and we have now reached a stage where a comprehensive body of knowledge, both theoretical and empirical, has been developed in this area. This book aims to provide practitioners, students, and academic researchers with a single point of reference on intermittent demand forecasting. Although there are considerable openings for further advancements, the current state of knowledge offers organisations significant opportunities to improve their intermittent demand forecasting. Numerous reports, to be discussed in more detail later in this chapter, indicate that intermittent demand forecasting is one