Free Subscription

  • Access 15 free news articles each month

Professional

Try one month for $5
  • Unlimited access to news,insights and opinions
  • Quarterly and weekly magazines
  • Independent research reports and forecasts
  • Quarterly webinars with industry experts
  • Q&A with retail leaders
  • Career advice
  • 10% discount on events
×

How to predict unpredictable demand: forecasting in times of crisis

How can you predict today what consumers will want tomorrow? In standard times, retailers rely on sophisticated algorithms that use past sales data to forecast demand.

These historical models, however, become useless when sales take erratic and unusual patterns, as happens during extreme events. 

This is the challenge for retailers today: they need to plan demand for an unpredictable future, and all they have to work with are months of biased or inapplicable sales histories.

Combining artificial intelligence and human insights

When sales histories are reliable, automated demand forecasting tools greatly enhance retailers’ ability to optimise their investments in inventory, down to the most granular level. But when demand is unstable, automations alone won’t produce accurate forecasts.

“During times of unusual sales patterns, you need to combine technological and human insights to achieve a dependable projection of demand,” says Martin Kleindl, who heads the development of replenishment and supply chain solutions at software development firm LS Retail

Here are three ways retailers can combine human intuition and machine insights to create reliable forecasts, and optimise stock coverage in periods of uncertainty.

1. Manually fix biased histories 

Most retailers have experienced extremely atypical sales in the past few months. Even as the crisis progresses, recent sales patterns are unlikely to repeat. Since the AI can’t (yet) read the news, it can’t know that spikes in sales of luxury watches and Birkin bags were due to a possibly short-termed “revenge spending” spree. Retailers therefore need to manually adjust biased histories. 

“When you combine the machine’s analytics capabilities with your knowledge of external factors, you obtain the most accurate predictions,” says Kleindl. “The best demand planning software therefore enables you to intervene, and manually fix abnormal histories so they are not considered during modelling. At the same time, the ideal software will also automatically detect extreme statistical outliers, and adjust sales histories accordingly,” Kleindl advises. “This kind of automation doesn’t replace manual adjustments, but reduces the burden considerably.” 

2. Improve the quality of predictions 

With festivals and celebrations repeatedly cancelled or postponed, the usual yearly ebb and flow in demand doesn’t apply this year. Retailers need to use their knowledge of external events, and their expectations of how changes to these events might impact sales, to adjust prediction modelling.

“If retailers expect that a specific product group will sell more or less than usual due to external factors, they need to be able to manually adjust the forecast. Then, the automations can take care of the heavy lifting, such as shaping a curve based on trend expectations, and adjusting quantities of items and variants,” says Kleindl.

3. Redistribute items across the chain

With both regional and localized lockdowns continuously eased and reintroduced, we should prepare for more forced store closures. To avoid being left with tons of valuable stock parked in locked stores, retailers urgently need tools that enable them to redistribute items easily and flexibly across the entire retail chain.

“The ideal replenishment software solution gives you the freedom to propose where items should be moved based on rules. This way you can optimise distribution to the stores that you know will be open and doing business,” says Kleindl.

“If you are looking for this kind of software now, go for one that also calculates automatically the time and cost of the different redistribution plans. The ability to move products quickly and cost-effectively will have a significant impact on the bottom line,” Kleindl advises.