So goes the perennial joke about forecasting. So, why bother then? Because a wrong forecast is better than no forecast
The fact is that businesses do forecast, because they have to. Where no forecast exists, then someone has to make one up. Otherwise, how do the people in purchasing know how much raw material to order, and how does the Managing Director know what to promise the shareholders ?
The danger of not having an effective forecasting process is that everyone is forecasting and every forecast is different. Purchasing will order more raw materials than production is forecasting to use; Inventory will plan more stock than the warehouse is forecasting to hold; sales will target less revenue than finance is anticipating.
What every business needs is one set of numbers with which every department can operate, developed using an effective forecasting process.
Forecasting is a process like any other. It has inputs and outputs, and it has owners. It is measurable, it has objectives, and it brings manifest benefits to the business. Better processes yield better results; better forecasting processes yield better forecasts. Forecasting performance is measured by accuracy, which goes back to the premise that the only certain thing about any forecast is that it will be wrong. However, wrong in what way, and how wrong? And, most important, what should be done to make things better?
Some degree of forecast error is inevitable. Inherent market variability guarantees this. What is important is to reduce error, and to focus on the most important types of error. First, it is forecast error, not accuracy that is important. Error implies the need for improvement without assigning fault. Without affecting the accountability of those that own the forecast, error acknowledges that there are a number of extraneous factors that impact demand and are often entirely unpredictable.
Second, focussing on forecast error makes it easier to attack the worst kind of forecast error: bias. Bias is the tendency for the forecast to err high or low. Over time, bias has a much worse effect on the supply chain than variability. Almost always, it is bias that causes overstocking or customer service failure.
Finally, it is a fact of life that it is easier to forecast large volume items than small volume items. Variability is less, and market factors are more easily predicted. Many items, however, sell in small quantities, and therefore are difficult to forecast. However, items almost always sell in families with similar characteristics, and can be grouped together for forecasting purposes.
This concept is the foundation of forecasting hierarchies, and of the philosophy “forecast less, not more” that is driving cutting edge thought on the forecasting process.
Developing a relevant and effective forecast hierarchy is core to good forecasting. It allows businesses to spend less time forecasting, it gives better understanding of markets and products, and it reduces forecast error. Forecasting using these techniques requires intelligent software that can apply flexible rules to convert high level forecasts to item level, without losing the subtle implications of lifecycle, substitution, promotions, seasonality, and many other effects.
It must also build a hierarchy that reflects the way your business operates within its markets, as well as allowing views of the data that do not conform to this. Such software must be capable of cutting the hierarchy in any direction, giving all functions their own view of the same data.
Much forecasting software relies on historical demand to make future predictions. But this approach only works if it caters for exceptions. Not only do things happen in the market, which buck the trend of a product’s demand, but the selling company often causes these effects. Obvious examples are promotions, stock takes, product launches and price changes.
Such effects must be catered for in any forecast. Not to do so is to court disaster, putting unforeseen pressure on the supply chain, or building up obsolete stock that can destroy the margins. However, forecasters should beware: managing such “events” means factoring them into the predicted demand and out of history; it means learning from them; it means logging them carefully so that the forecast can be correctly interpreted; and it means using a system that can apply events simply and rapidly.
Confidence in your forecast is vital to running an efficient business. The right forecast means the right inventory policy, and the right production or replenishment plan. One cannot exist without the other.
Producing an effective forecast, with all events and promotions catered for, allows a business to reduce the protective buffer it holds in inventory, and to improve service where stock is misbalanced.
The benefits of this strategy are immediate: it will reduce inventory levels, obsolescence, warehouse footprint and capital investment; it will improve service, cash control, order fulfilment and brand reputation.
Forecasting is something we will always have to live with, even if factory performances improve so dramatically that production lead times sit well within the customer demand cycle. We will still require forecasts to support budgets, shareholder reports, purchasing schedules and much more.
Making best use of the forecast requires understanding its limitations on the one hand, and applying it effectively on the other. This means using it as a vital input to the Sales & Operations Planning Process, and integrating the forecast properly into other systems in the business. Inventory optimisation, factory planning, DRP, demand management, available to promise, materials planning, and P&L are just some examples where the forecast can be used effectively to drive improvements in business performance.