The saying that seems to be most familiar to people studying economics and attempting to build forecasts is that “the past is the best predictor of the future”. Some people will add to it that this is the only information we have.
Thus, it will not surprise you that this is also the basis for most forecasting models used in retail. However, in today’s erratic environment it has become increasingly difficult to predict what consumers will want to buy in the future.
Forecasting is difficult in the best of times. The algorithms that use historical sales models and adjust them for seasonality to project future demand have error margins. In crisis times these may sink the ship. They break down when faced with the uncertainty of a pandemic. In the long term, airlines who have very sophisticated pricing algorithms, may show retail a way to make sense of the crisis. However, in the short term, we may have to help ourselves using a few different tools.
Automation Tempered By Humans
Demand forecasting tools have made their way from larger companies into many medium size and small retailers. These tools are extremely valuable and can help to improve your business’s bottom line. With good and clean history of data, forecast can be very accurate. However, when we have the current market volatility we cannot rely on these tools alone. Human insights have to be combined with the analysis provided by these tools. This is not just a software play, but also a play using experience and sense-of-business. In short, combining science and art.
Is this all that is needed?
Substantially we will answer the question with “Yes”. But, you can do more. And, in the shorter term, with a constantly changing situation and even more lock-downs ahead in the future, we will suggest a few other approaches to help you make better forecasts even when encountering irregular demand patterns.
Clear the Exceptions
For the majority of retailers the COVID season was atypical in sales pattern and is not likely to be a repeating pattern. Even as the crisis continues the data gathered during the crisis is going to be impacting your future forecasts unnecessarily. Computers, at this time, are poor interpreters of current these exceptions. For example, a current spike of leisure ware is not going to continue after COVID when workers return to offices.
As retailers this means you have to look for bias in your data and clear the exception queues. You need to detect statistical outliers and adjust your sales history to reduce their impact. The human knowledge of external factors has to be combined with machine analysis to arrive at better retail predictions.
Emphasize the Human
With both regulations and shopper behavior changing rapidly, neither recent sales patterns nor historical sales trends can be used as a basis for a trustworthy forecast. In times of high volatility, the retailer’s sales expectations, based on their knowledge of future events – the delay in back-to-school plans; the cancellation of weddings or of Halloween celebrations – are tremendously valuable, and should be used more heavily to model and adjust predictions.
Once the human directed inputs are clarified the software can be used more effectively to make predictions. But, even in that pattern, the ability to respond quickly should be considered and appropriate trend expectations should be built into the forecast models.
Re-balance your Supply
The pattern of lock-downs and easing can impact your inventory availability in unexpected ways on the regional and local level. Retailers should prepare for more forced and unexpected store closures, which may occur with little to no advance notice. There’s a real risk of having valuable stock stuck in locked stores, which makes it both important and urgent to secure tools to redistribute items easily and flexibly across the entire supply chain.
This is another balancing act that may be supported by your software. Or, you may need to look for additional tools that can help you model store closures and retail supply redistribution. Thus, your software solution should be able to accept modeling rules with constraint of retail availability so you can use knowledge of external factors to create an optimal distribution for the times of crisis. With the time and cost for each redistribution plan more clearly outlines you can make better decisions for your supply chain overall.