As we’ve previously discussed here at the APEX of Innovation, the pandemic has thrust global retailers and brands into new and challenging waters. Traditional demand forecasts were essentially rendered useless, and much of this unpredictability persists today. In addition, widespread unemployment and the volatile economic and political climate are introducing retail uncertainty. Finally, consumer buying behaviors have likely been fundamentally altered after all we’ve experienced in 2020.
A recent article from the Harvard Business Review (HBR) cautioned that, when faced with issues such as those outlined above, brands often make things worse by reverting to their gut instincts and, “confusing noisy data with bias.” This can lead to a host of issues, including either an excess or dearth of inventory, damaged channel relationships, and lost sales.
So what can brands do to predict demand when much of the industry remains unpredictable? The HBR piece offers the following suggestions:
- Find Alternative Data Sets: The authors urge companies to examine data from previous economic shocks, particularly hurricanes or other natural disasters in which supply chains were significantly disrupted. These analogous events can hold predictive power and also help brands determine how to structure their ongoing COVID-19 response. Additionally, brands may look for ways to get data from retailers more efficiently or consider building direct-to-consumer channels if the latter proves difficult.
- Tap Local Knowledge: Local viewpoints are critical, particularly given the regional variances in the pandemic response. For example, the 2020 back-to-school season varied greatly depending on location; understanding these changes and their impact on product demand is only possible by talking with local reps who have first-hand knowledge. Companies may also want to speak with epidemiologists or others involved in the COVID-19 response. These expert opinions can then be built into the data sets that go into building models, increasing the accuracy of the ultimate output.
- Embrace Ensemble Modeling: According to the article, in dynamic conditions such as our current climate, combining many simple models often works better than using just one complex model. Ensemble modeling draws on predictions from different models to recommend a reasonable range when the underlying data for any single model is unstable.
- Test, Test, And Test: Consumer demand is fickle, particularly in challenging times. The pandemic has changed buying patterns and also forced more people online. For example, many older adults likely bought groceries online for the first time when lockdown restrictions were announced. In this environment, A/B testing can be used to assess the effectiveness of various marketing channels, which marketing messages are resonating best, and other metrics that can help companies reach consumers more effectively.
For more on these and other strategies, check out the HBR piece in its entirety here.