Inventories as work backlogs-Three case studies in diagnostic analyses using regressions, moving-averages and cum-sum charts

The first case study in this paper deals with a concern for detecting whether the inventory at any point in time is significantly due to a slackening of work at the downstream facility. It uses the well known cumulative-sum chart for this purpose, drawing upon the analogy of a random walk for the inventory process. This predominantly mechanistic view of the inventory nevertheless proved useful in the case situation in the receipt stores of a large factory where it was feared that workers were sometimes accumulating backlogs to earn overtime later. The second case study, also in the same company, concerned issue of materials against requisitions. Here it was statistically discovered, or rather conjectured first based on a certain regression and later confirmed, that work backlogs were often created by workers practically ignoring the material requisitions until the day before the due date for materials’ issue. Thus an important cause for some very expensive and crippling delays was discovered statistically. This case also brought into light the effects of the fact-of-life on the plant floor; that the output rate of the human work station is sensitive to the work backlog as seen by the work group. Ignoring work until its due date itself meant under-estimating true work backlog and thence slackening of work output. Whereas the first two cases dealt with situations arising out of workers partly postponing today’s work till tomorrow, the third case study, interestingly also in the same switchgear plant, deals specially with the consequences of workers stretching work when work backlog is low and compressing or accelerating work when work backlog is high. An argument essentially concerning propagation of errors, together with use of regression of work output rate upon work backlog, yields a representation of work backlog in terms of a moving weighed sum of deviations of work input rate (in discrete time) from mean input rate. This modelling positively enabled the management to set objective norms for work backlogs instead of setting them primarily by subjective hunches or negotiations with labour representatives in the past.