The Right Level at Which to Plan

It is always a struggle to get it right: at which level should you forecast to obtain maximum accuracy for your demand plan, but without spending too much time in the detail. Is there an easy solution?

Not really. It’s really a blend of effort vs reward. Do you spend your time improving the forecast of 3-5% when the sales involved represent less than 5% of the total sales? Probably not, but the following considerations may help you to decide whether to combine or forecast a lower level. It’s ultimately about getting the most accurate forecast at all levels, not just at that level.

There are generally two dimensions within a demand forecast – Product and Customer. Each of these have a hierarchy, or a set of attributes that can allow you to segment these product/customer combinations into groups to better manage forecast. In addition, there is a time element, and believe it or not, there is also a fourth dimension: Sales Behaviour.

Sales Behaviour

Behaviour of the product within certain channels can be the most influential in your ability to forecast. Cutting corners here can result in some poor accuracy results. Areas to consider are:

·        Seasonality pattern – if you’re combining sales together where the actual seasonality is different, you will get inaccuracies.

·        Frequency – regular sales combined with irregular sales patterns can result in a false seasonality pattern

·        Growth Factors – combining sales showing significant growth with sales showing the opposite will dampen the effect of both, resulting in lack of stock on the growth element, and too much stock in the diminishing element.

·        Product Lifecycle – combining a new product/channel into an ongoing sales pattern may again result in the wrong sales signals being applied to the forecast.

Combining all four of these elements together, (as they are never found in isolation) can result in many forecasting inaccuracies.

Periodicity

The decision to forecast in weeks, or months or even days depends upon the cadence of the supply as well as the sale. If, for example, your purchasing cadence is monthly with stock arriving once a month, consider why you would you need to forecast at a weekly level? If, however you’re selling daily, and receiving supplies weekly, a weekly periodicity may still be too small, depending upon the nature of the product. Ensure the periodicity of the forecasting application matches the needs of both supply and demand.

Product Characteristics

My experience working with customers, has identified some characteristics worth considering with products before you consider combining them all together:

·        Colour – at a FMCG business I was recently working with, we were able to ascertain that white products’ sales patterns varied differently to their black offering. Forecasting them separately resulted in improved forecast accuracy.

·        Size – believe it or not, different sizes of the same product can behave differently.

·        Product Usage – another FMCG customer we recently worked with was showing different growth patterns for a specific product within a range of products designed for outdoor use. After further investigation, it was found that sales of the product were also being used indoors, in an application not hitherto considered. By separating that product out, they were able to increase their forecast accuracy for the product by over 20%.

Customer Characteristics

When you look at the customer dimensions, there are more things to consider:

·        Sales Channel – consider the channelization. Sales into a Duty-free channel behave differently to a store in the centre of a large city. The seasonality differs – duty free has spikes when you wouldn’t necessarily have expected.

·        Industry – sales of the same product into one industry can be different from another. Different industries can have different seasonality patterns and also different growth patterns.

·        Geography – Northern hemisphere customers act very differently to Southern Hemisphere. This is not just the seasonal difference, but also trends: finish, size, colour trends can all be different, and this can also be apparent in more localised differences between Northern and Southern Europe. (Did you know there’s a 12 cm difference between the average male height per country across Europe?)

How and Where Is The Forecast Going To Be Used?

On the assumption that the forecast will be primarily used for supply chain planning, and ultimately in the purchase of resources, other business units are likely to be using it for their own needs. This should be encouraged, giving the business a one-truth view of the data. Finance, Sales Management, Marketing, they may all have a vested interest in the forecast, and the levels at which they require it may have a bearing on the level at which you forecast, and this could mean using data at a level which is lower than is optimum for creating a demand forecast.

A good demand planning application will allow you to load data at a base level and create forecasts at a higher aggregated level to gain the most accuracy, to then be shared with the different business units at the level they need it at.

Base Level vs. Aggregated Level and Segmentation

Additionally, you shouldn’t think that the level to forecast should be the same for every sector of your business. If you are getting good accuracy at customer level in the UK, but better at product series level in Spain, segment at these different levels in different markets to get the most accurate forecast.

Most of the best demand planning applications on the market today allow you to run forecasts at different levels within a sales hierarchy in order to get the most accurate of forecasts, which then populate the forecast at all other levels within the hierarchy. With this ability, you can analyse and choose the best forecasting algorithms for the right levels to give you the most accurate forecast.

One of the features that Demand Solutions customers find really useful is the ability to run forecast algorithms at different aggregation levels, compare and contrast between them to choose the most accurate demand plan.

Conclusion

Demand Forecasts within an S&OP process drive everything in the business, so it’s important to get the demand forecast that is as accurate and effectively created. Time shouldn’t be scrimped when considering the importance of segmentation and aggregation, but it is an often-missed area of demand forecasting. Ultimately, it’s about getting this blend of characteristics right for your business, to get the demand plan right.