Question 7 of 7: Materials Handling, Inventory Control and Forecasting Model Selection

Question 7: Materials Handling, Inventory Control and Forecasting Model Selection (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Part (a) — objectives of materials handling. Materials handling adds cost and risk but no value: a part is worth no more after it has been moved than before. The governing objective is therefore to minimise handling while ensuring that material is where it is needed, when it is needed, in the right quantity and undamaged. That general aim resolves into a set of specific objectives.

The first is reduction of handling cost, which in many plants is a quarter or more of total manufacturing cost, by eliminating moves outright, by shortening those that remain, by combining moves so that material travels in larger unit loads, and by mechanising the moves that cannot be eliminated. The second is reduction of throughput time and work in process: since material spends the overwhelming majority of its time in a batch shop waiting or moving, cutting the handling network shortens lead time and releases working capital directly. The third is better space utilisation, both floor and cube — handling equipment determines aisle width, stacking height and the size of staging areas, and poor handling design consumes floor that could be production space.

The fourth is safety, and in Canada it is a legal duty rather than an aspiration: manual handling is the largest single source of lost-time injury in manufacturing, and provincial occupational health and safety regulations require that risks of musculoskeletal injury be assessed and controlled, with mechanical handling being the preferred engineering control. The fifth is protection of the material: damage, contamination and mixing of lots all originate in handling, so containerisation, unit-load standardisation and correct fixturing during transport are handling objectives, not packaging afterthoughts. The sixth is improved control and traceability: a disciplined handling system with standard containers and identified locations is what makes accurate inventory records, shop-floor tracking and first-in-first-out rotation possible at all.

The seventh is support of the production system's flow. Handling equipment should be chosen to suit the flow the plant wants — conveyors and fixed paths for high-volume repetitive flow, forklifts and tuggers for variable job-shop routings, AGVs and rail-guided vehicles for flexible manufacturing systems, robots and gantries where the handling is part of the machining cycle. Finally, the whole must satisfy flexibility and economic justification: handling equipment outlives the products it serves, so it must be capable of adapting to changed routings, and each installation must be justified on the cost per unit handled, not on its sophistication. The classical principles — plan the whole system, standardise, use unit loads, use gravity where possible, minimise dead-heading, keep the flow direct and short — are simply these objectives expressed as design rules.

Part (b) — what an effective inventory control system should accomplish, and the vital areas. An effective inventory control system must accomplish six things. It must meet the service objective: supply the material required by production and by customers at the agreed service level, which is the reason inventory exists. It must do so at minimum total cost — the sum of ordering or set-up cost, holding cost, shortage cost and unit cost, traded off against one another rather than each minimised alone. It must maintain accurate records of what is on hand, on order and allocated, because every decision the system makes is only as good as the balance it starts from. It must answer the two operating questions explicitly for every item — how much to order and when to order — through a stated policy rather than through the judgement of whoever is on duty. It must give visibility and exception reporting, drawing management's attention to the items that are out of control rather than reporting on all of them equally. And it must protect the investment: identify obsolete, slow-moving and excess stock, and control shrinkage.

The vital areas to consider when developing a comprehensive system are the following. Classification and selective control: an ABC analysis by annual dollar usage, so that the ten or twenty per cent of items carrying most of the value receive tight continuous review while C items are controlled by simple two-bin or periodic rules — without this the system spends its effort in the wrong place. Demand forecasting, since every reorder point and lot size is built on a forecast, and the forecast error, not the forecast itself, sizes the safety stock. The cost structure: realistic ordering and set-up costs, a defensible carrying rate (capital, storage, insurance, obsolescence, taxes, typically twenty to thirty per cent per year), and an explicit view of what a shortage costs. The lot-sizing and reorder rules themselves, including which items are on continuous review and which on periodic review, and whether dependent-demand items should be controlled by MRP rather than by statistical reorder points — a distinction that is frequently missed and which invalidates reorder-point logic on assembly components. Lead times and supplier performance, both their average and their variability, because variability drives safety stock more strongly than length does. Service-level policy, stated by item class as a management decision rather than left implicit. Record accuracy and physical control: cycle counting, locator systems, controlled stores access, and reconciliation procedures. Systems integration with purchasing, production scheduling, cost accounting and sales. Performance measurement: turnover by class, fill rate, stockout frequency, obsolescence write-offs, record accuracy. And finally the organisational questions of who owns the decisions, who may authorise an exception, and how the policy is reviewed as demand and costs change.

Part (c) — factors influencing the selection of a forecasting model. The choice of forecasting method is governed by the following factors.

The time horizon. Short-range forecasts of days to a few months support scheduling and inventory replenishment and are best served by time-series methods; medium-range forecasts of months to two years support aggregate planning and are served by decomposition and regression; long-range forecasts of years support capacity and facility decisions and are usually qualitative or econometric. No single model spans the range.

The pattern of the data. The series must be inspected for level, trend, seasonality and cycle before a model is chosen. A stationary series is served by a moving average or simple exponential smoothing; a trended series requires double smoothing or Holt's method or a regression on time; a series with both trend and season requires Winters' method or explicit decomposition; irregular, lumpy demand may need Croston's method or a dependent-demand approach instead of a forecast at all.

The availability, quantity and quality of historical data. Time-series methods need enough clean history — several full seasonal cycles for a seasonal model — and a new product has none, which forces a qualitative or analogy-based method until history accumulates.

The required accuracy, and the cost of error. A high-value item whose shortage stops an assembly line justifies an expensive method; a low-value C item does not. Accuracy is measured on holdout data by MAD, MSE, MAPE or the tracking signal, and the model that fits history best is not necessarily the one that forecasts best.

The cost of the method itself — development, data collection, computation and, most of all, the analyst time to maintain it — weighed against the saving the improved accuracy produces. The number of items to be forecast matters for the same reason: a system forecasting fifty thousand stock keeping units nightly must use a method that runs unattended and self-adjusts, which is why exponential smoothing with adaptive response dominates inventory forecasting.

The stability of the environment. Where the underlying process is stable, extrapolative methods work; where it is being disturbed by promotions, a competitor's entry, a regulatory change or a technology shift, causal or qualitative methods are needed because the history no longer describes the future. The availability of causal variables follows from this: a regression or econometric model is only usable if the driving variables are themselves known or forecastable in advance.

The user. A method whose logic the planner does not understand will be overridden, so simplicity and transparency have real value; and the form of the output required — a point forecast, a range, or a full distribution for safety-stock calculation — may itself rule methods in or out. Finally, the level of aggregation matters: forecasts of families are far more accurate than forecasts of individual items, so the model should be applied at the highest level of aggregation the decision permits, and the result disaggregated.