Traditionally, when forecasting for events and promotions, retailers simply took the sales from the last x years, combined that data with current sales information and calculated a plan for the upcoming event. In the days of a single-channel environment with more predictable shoppers, that system was fine.
However, that is no longer the world in which we live. Ever-emerging channels, the battle for loyalty and the rise of the digitally informed consumer have rendered the old methods completely obsolete. Add to those issues the unexpected and dramatic disruptions that have occurred with health issues, weather and economic shifts that we’ve endured over the last couple of years, and it seems impossible to predict what’s coming next.
In the fifth installment of my series on demand forecasting, I’ll be talking about why retailers need to be more agile when it comes to events and other disruptions.
The Question: Can you accurately forecast for promotions, seasonal events and weather and share these forecasts with suppliers to improve availability?
Events, holidays, and the promotions that typically accompany them are complex. What makes them even more complex, from a forecasting perspective, is that, after the events of 2020, it’s practically impossible to decipher historical demand that is inflated or distorted due to running out of stock – which may include demand events or promotions or may not. Traditional forecasting systems depend on recurring events to forecast the future, and don’t understand anomalies in demand or fluctuations in events so this presents a fundamental problem.
Traditional forecasting systems deal with the problem of “lost past” by grouping like items together and making assumptions such as, two similar products in the same category will behave the same way. Making these types of assumptions greatly oversimplifies the actual demand for the various items in a given category and can lead to out-of-stocks for one product and overstocks of another. Retailers need to be able to assess the nuances of all products to determine the best levels for all dimensions – store, cluster, promotion, event, and channel.
Then we come to the forecasting issues specific to promotions. In some cases, the supply chain may not even be aware of a promotion, leading to out-of-stocks and a potential detriment to shopper loyalty. Even when the promotion is known, the supply chain may not know which mechanics have been applied and how the promotion needs to roll out and be fulfilled. Lastly, the supply chain needs to be able to monitor whether the promotion execution was carried out properly across all channels and, if not, what specifically happened to reduce the effectiveness of the promotion.
An inaccurate forecast is bad enough…
But what happens when you go share the forecast? In an ideal supply chain, a retailer would build a demand forecast built on a single version of the truth. That forecast would then be shared across the business and with suppliers to execute based on desired outcomes and align with end-to-end supply chain objectives. However, when forecasts are built with mistakes or omissions it causes a ripple effect throughout the retail organization and its partners resulting in accurate plans, poorly executed promotions, and strained relationships with suppliers.
Today, retailers and suppliers must be more aligned than ever to ensure that any disruptions in the supply chain don’t result in the issues we endured in 2020. However, if forecasts for promotions cannot be effectively and accurately shared with suppliers, disconnects will occur and the risks of over or under-stocking for the promotion grow exponentially.
Be certain that your prospective demand forecasting solution can respond to any combination of recurring events – such as school breaks, holidays, sporting events – and unexpected disruptions – such as weather, economic and health events – in an automated way, removing human intervention and vastly increasing accuracy.
AI plays a critical role by recommending the best possible forecasts for events – expected and otherwise – based on more robust modeling, machine learning, and business rules. Because retailers have billions of data and decision points to consider, AI is truly the only way to seamlessly and quickly address complexity by identifying patterns to analyze and diagnose complex problems with accuracy that far exceeds human capabilities.
The benefits of AI and machine learning:
- Machine learning takes forecasting to a new level – It constantly updates the models in real time with additional data feeds that continuously refine the forecast accuracy and optimize trade spend.
- AI performs many of the backend/administrative tasks typically done by individuals (or worse that are not currently done because of the amount of time and effort it takes today). For example, AI continuously monitors internal and external data sources for signals that provide more granular alerts or prescriptive recommendations at an item or store level – empowering retailers to course correct more quickly.
- By providing a single version of the truth in near real-time, AI enables retailers to confidently share accurate and up-to-date forecasts across the enterprise and with suppliers. This results in faster reaction times to shifts, better executed promotions, higher levels of collaboration and stronger relationships across the entire supply chain.
Read part 1 of the series: Is your demand forecasting solution working for you or are you working for it?
Read part 2 in the series: Are your demand forecasting systems connecting all categories across the store?
Read part 3 in the series: Why demand forecasting must be effective with all channels to be effective with any
Read part 4 of the series: Avoiding the ‘blindside’: Inventory visibility and the agile supply chain