Forecast for Success

There is an assumption in the market that more forecasting models must be better. The correct question is not how many or which models, but rather what is the right methodology for the business problem you need to solve.

Are you a manufacturer making goods? Or are you a wholesale distributor or retailer who replenishes goods to your locations for sale to other businesses or consumers? These are significantly different business problems and Blue Ridge addresses both individually.

If you are a wholesale or retail business contemplating better ways to manage your inventory investment, you should be thinking about inventory as one holistic investment for your enterprise. Your primary goal should be to enable the most profitable inventory investment that meets your customer service objectives. The primary business user is a combination forecaster and inventory optimization analyst. Different businesses may label them differently. Whether it’s a replenishment analyst, replenishment buyer, or something else, these are the people who are investing your company’s money in inventory to ensure service to your customers. This is best accomplished using forecast and replenishment/inventory optimization software that is designed from the ground up for the profitable replenishment of finished goods in wholesale distribution and retail.

If you are a manufacturer, forecasting drives your manufacturing requirements planning (MRP) process and raw material sourcing. Using a software application for this business problem requires one or more full-time forecast analysts (we haven’t gotten to purchasing yet) who work with demand time series streams to prepare for a forecasting operation. Often, manufacturers don’t own the demand stream from which they generate forecasts. A manufacturer’s business problem is primarily the effective sourcing and production of the goods they make.

One of the big differences between wholesale and retail compared to manufacturers is how and why demand forecasts are developed.

Differences in Forecasting

Manufacturers have to generate demand forecasts to drive their requirements planning and material sourcing.  There are many differences in why and how a manufacturer generates demand forecasts versus the wholesale and retail community.  These differences impact forecasting in several ways:

Time Series Demand Stream

Most manufacturers do not inherently ‘own’ the demand stream from which they generate their forecasts. Rather, they get disparate types of demand, sometimes even once or twice removed from them, that they have to accumulate into a workable format to perform a forecasting operation. For example, they may have shipments from their own DC, plus numerous point of sale signals from downstream customers that have to be combined and reconciled before they can generate forecasts.  This introduces a dangerous level of subjectivity, but it must be confronted nonetheless. That means their process has to accommodate scenarios wholesalers and retailers would not encounter.

On the other hand, the wholesale and retail community is much closer to the demand signals from which their statistical forecasts are generated. They also understand spikes, outliers and other data signal anomalies. It doesn’t mean one approach to forecasting is better than the other. Various business problems require specific forecasting solutions.

The Forecast Analyst

Manufacturers’ forecasting processes are predicated on the idea of having dedicated forecast analysts who strictly work with demand streams and forecast models. This is the nature of their business. They are working with relatively small lists or catalogs of items compared to the smallest of wholesale customers who are working with thousands of items.

SKU volume is vastly different between manufacturers, wholesalers, and particularly retailers. The forecast analysts in manufacturing operations are essentially completing a forecast initialization each time they “update” a forecast.

Forecasts Have Different Downstream Implications

The manufacturing business model requires longer-range demand forecasts. The forecast that starts out as statistically derived, is pretty much the only thing to go by. This is because their lack of immediate access to consumer/SKU/location-level data forces them to make assumptions that are not required of a wholesaler or retailer. Then, they add in their own demand-shaping techniques like promotions and collaborative tribal knowledge (speculation).

Wholesalers and retailers, if interested in optimizing profit, need the demand forecast to drive economic order quantity replenishment. Generally speaking, wholesale and non-fashion retail are working with reasonably long product life cycles, relatively short to moderate lead times, and unlike manufacturers who have fixed cycle buying/pre-determined production schedules, short order cycles and variable order schedules.

Wholesale and retail need more frequent forecast updates because the forecast is a critical path input to the daily update of their inventory investment. It is a significant part of optimized safety stock and economic cycle stock. In wholesale and retail, the inputs to these components change. There is no need for a person to touch every forecast. Wholesalers and retailers need to concentrate on monitoring their inventory investment and allow the solution to work on forecast updates. To do this optimally, wholesale and retail clients need a unique synthesis of a frequent, adaptive forecasting model coupled with economically-optimized safety stock and cycle stock.

Blue Ridge wholesale and retail customers can do long-range planning effectively because they have better tools for finished goods. Our joined aggregate, time-phased order projections take into account not only the statistical forecast but also other segments of the forecast such as causal variables and strict buying policies including the manufacturers buying multiple, minimum order sizes, and so on. This is all part of the forecasting process for wholesale and retail customers - much more than just statistically derived forecasts. This forecasting method is also a valuable long range planning input to share with manufacturer partners. Sharing what they will order in the future is much better than simply sharing point-of-sale or some other demand signal.

Multiple Forecasting Models & "Best Pick/Best Fit"

When you are practically reinventing the forecasting wheel every time you generate forecasts, much like forecast analysts in the manufacturing world ( i.e. they have to manage and reconcile time-series demand signals and they are not necessarily intimate with the dynamics of the various demand feeds), it is incumbent on the forecast analyst to try different models to see which gives them the lowest forecast error or have the solution change the forecast model based on breaching an error threshold. If the solution does it for them, that is what is referred to as “best pick or best fit” forecasting.  Again, typically they are consistently doing much longer range production planning, such as quarters or even years into the future.

Wholesalers and retailers need the SKU demand forecast to be smoothed and updated more frequently based on the latest trends, utilizing a rapidly adaptive model that uses a combination of proprietary extensions on multiple forecast models designed to adapt to the item/location demand behavior in wholesale and retail inventory optimization.

Conclusion

The myth that a solution originally designed for MRP can work for wholesale and retail supply chain planning introduces confusion. This is especially true for wholesale and retail businesses when trying to ascertain which forecasting solution is best for the unique demands of their business.

Blue Ridge Supply Chain Planning customers already have it covered.