The secret to better forecasting?… it’s classified…
Deciding the right balance and assortment of stock-keeping units across multiple warehouses can be a bewildering task. Regularly monitoring and planning the correct level of inventory across tens of thousands, or even hundreds of thousands of items without the right tool is enough to make even the best of supply managers capitulate. Adding the difficult task of deciding which forecasting technique will produce the most appropriate forecast results for a given demand profile, along with all of the configurations available for each model, results in endless possibilities and permutations and a very challenging problem indeed. But what can supply chain managers do to overcome this hurdle?
Demand Classification is a systematic, logically founded method of ascertaining and clustering the characteristics of sales history into groups, before applying the most appropriate forecasting techniques automatically. Classification enables a significant improvement in forecast accuracy with significantly less effort and is light-years ahead of the traditional ‘pick best’ approach.
New, in-life, seasonal, and end of life products must be all treated differently, and most demand patterns change. New products may move from a growth phase into the mainstream, lose their upward trend and become more stable. Others may exhibit stable demand patterns and begin to trend downward toward the end of a product’s lifecycle. Along with sales velocity, frequency, and magnitude, demand can be different in every case. Planners quickly begin to recognise the daunting task of just knowing where to start, followed by where to focus their priorities. Not only are forecast accuracy goals negatively impacted by their follow on action and decision, but downstream activities can be adversely affected too. Very quickly inventory and replenishment planning begins to suffer and an extremely costly situation ensues.
With so many planning decisions pending a forecast, many practitioners have been tempted to rely on their computer system to select, for them, the best prediction model from a wide range of techniques. Compared to more scientific approaches, this might only be a little more advanced than the random movements of a laboratory rat seeking to exit a maze. The forecast models in a pick best approach can produce wildly divergent results. Different samples of history will influence predictions, and performance is often negatively impacted by significant fluctuations in the data. The technique usually creates endless hours of model tweaking and data manipulations to gain only slightly less dubious confidence in the forecast. Fitting a forecast based on how tightly it fits the past, does not guarantee a better alignment to the future, and it’s very often the root-cause of kneejerk planning.
A more rational approach to the problem is to include statistical recognition of different types of demand before precisely applying the right technique. Failure to recognise the differences in demand characteristics can result in the application of inappropriate planning techniques and create undesired fallout like excess inventory, wild swings in replenishment plans, late promotional deliveries, and more frequent stock-outs. Slimstock’s Slim4 Inventory Optimisation software automatically recognises what stage of lifecycle a product is in, and systematically analyses demand history or replenishment challenges before scientifically applying the optimal planning strategy.
Slim4 seamlessly integrates with any ERP system and applies academically supported statistics to test trends across each product/location combination, identifying the appropriate classification for items that contain seasonality, irregular or lumpy demand, new activity, declining sales, and many other observations. This technique dramatically reduces the amount of effort from traditional forecasting and planning techniques alone, allowing planners to engage in additional activities that drive even greater business improvements. The result is more accurate and consistent forecasts, and inventory replenishment plans based on scientific intelligence, with far less effort. Demand Classification is a significant leap forward from the traditional ‘pick best’ approach. The obvious benefit is a dramatic improvement in the overall confidence of an inventory planning strategy.
At Slimstock, we believe a significant opportunity exists in releasing a tremendous amount of kinetic energy (cash-flow) from the inventories of most companies. The way we help our clients capitalise on this idea is designing software that becomes the ‘central intelligence’ of your inventory planning organisation and helps pre-empt threats on your supply chain objectives. At Slimstock, we don’t just replicate ideas; we are advancing supply thinking and helping our clients achieve extraordinary results from near ‘mission impossible’ scenarios.