Simplify Demand Forecasting
Paul Trudgian Ltd | Supply Chain & Logistics Consultancy No Comments

Despite technology bringing with it improved clarity and visibility in our supply chains – through data acquisition – much of a supply chain’s understanding of supply and demand still comes down to accurate demand forecasting. How do we improve the accuracy of our forecasting? Indeed, how accurate can we expect our demand forecasting to be?

Demand Forecasting – Science or Art?

Forecasting is, in many ways, part science and part an art form. It is definitely a skill. In order to reduce errors in forecasting we need not only to have accurate data, but this also needs to be paired with experienced intuition. Knowledge of data and statistics is invaluable, but with it needs to be paired industry-specific experience in order to understand customers, products and markets.

Improving forecasting accuracy is essential because of its notable effect on the bottom line. If forecasting can be improved then there will be greater efficiencies, and cost reductions, when it comes managing inventory, manufacturing, logistics and more. Alongside this, improved accuracy in forecasting will improve customer relationships, as you are better in tune with demand and creating increased satisfaction.

However, forecasting is still unlikely to be 100% accurate. In addition, some will argue that Just-in-Time deliveries make forecasting worthless, but this doesn’t fully understand the real picture. Even with a highly-refined and short manufacturing lead time, which doesn’t exceed customer expectations for order fulfilment, you still need forecasting to help you plan and prepare components, parts, and raw materials.

Different Methods of Demand Forecasting

When looking at the supply chain industry as a whole, it is possible to identify the most successful methods for improving accuracy in forecasting.

Trends can be analysed, and these alone will help forecasters in many niches. Many within forecasting roles will be all too familiar with utilising Modified Holt, Double Exponential Smoothing Technique, and Holt-Winters Triple Exponential Smoothing Technique, to refine forecasting accuracy. However, these methods are not always applicable, especially when your products don’t seem to have a clear demand pattern.

In these cases, forecasters need to consider more straightforward averaging techniques, such as the Naïve Forecast Method which uses previous results. Whilst this is often seen as a crude forecasting method, it can be highly effective. In other situations, a Kanban system may be best, especially where items are either low in value and volume, or which are easy to get hold of.

However, all of these methods of forecasting still struggle when there is intermittent demand.

Looking at past performance can give no clear indication of what is to come. The Modified Croston Method can bring some forecasting sense to such demand which has minimal or no pattern. This method aims to bring regularity where none readily appears.

We see a similar problem when a new product is coming on to the market. Forecasting becomes extremely tricky. In many ways this can seem to be a specialised area. However, in reality, few products are completely and unequivocally ‘new’ – they usually share traits and similarities with other products already in existence. It can therefore be prudent to forecast demand using other products and brands already in existence.  This format of forecasting is known as Derived Modelling. It utilises existing past data for similar products to forecast for the new one. The success as a forecasting method is largely dependent on choosing the right product to work from in the first instance.

An alternative to Derived Modelling is the Attribute-Based Model. This is another option for forecasting new product demand but can also be used for those products believed to carry a short-life cycle (perhaps for an event for example), or for products which don’t as yet have a long history to draw on. Similar to the Derived Model, a product is assessed according to its similarity with others in terms of elements such as functionality, brand, material or size. From these a demand profile is created according to specific attributes. The most appropriate profiles are then used to forecast demand for the new product.

Using advanced technology, it is possible to automatically create suitable profiles, but to also match new products to them. By introducing more computer intelligence to the approach it is possible to monitor forecast accuracy and improve it over time, as well as in ‘the moment’ according to new data.

Still looking at Attribute-Based Models, there is a specific type called Proportional Profile Planning (PPP). This takes disaggregate higher-level forecasts for various selections of products and makes them usable as lower-level forecasts. PPP has gained immense popularity and success in the footwear industry and is used on a mass scale. PPP has become a favoured choice for many because it involves less effort and input, but brings with it improved accuracy.

The last forecasting modelling option which may prove valuable, most notably where promotions and advertising campaigns are brought in to the mix, is the Causal Model. The idea is to create relationships between different variables, in this instance between the product’s demand, and the promotion. It should help to forecast the correlation between the two. A specific type of Causal Modelling is known as Neural Network Forecasting, which effectively ‘learns’ through a process known as ‘backcasting’ so that it can predict and determine future promotional plans.

Forecasting Demand in the Modern Supply Chain

The modern supply chain is more than ever dependent on accurately forecasting demand. With computer modelling, we have more tools readily available. It’s a mixture of knowledge, expertise and experience to know which model to use when. It is also a matter of competitive edge and meeting customer expectations to select the right one. Therefore simplifying forecasting is not straightforward, but instead is a process of combining talented individuals, the right technology, and the rights models, to get the best results.

If you need advice on forecasting demand in your supply chain, why not use one of our specialised supply chain consultants to help? Call today on 0121 517 0008.

Leave a Reply

Your email address will not be published. Required fields are marked *