Without reliable data, forecasts can’t be created. Without human intuition and experience, however, companies with reliable forecasts might make bad decisions. Therefore, striking up a strong, flexible relationship between data, analytics and those who interpret the results enables companies to tackle even the trickiest forecasting challenges.
And there are a few.
What do you do when there is no history from which to forecast? Perhaps you're introducing a new line of products and have no past sales data to reference. Or perhaps you're launching a new store. Perhaps you're doing both!
If you lack specific product data, good forecasting analytics will borrow it from other, similar products. A new line of soft drinks, for instance, is more likely to behave like an existing soft drink than, say, tea or coffee.
If you're launching a new store but have other stores, good forecasting analytics will answer the question: which of these stores most closely resembles the new store?
What do beetroot juice, ginger shots, and protein pudding have in common? They went from obscurity to top sellers before settling at lower, stable normal, all within the space of 18 months.
Whether it’s an individual item or an entire segment that is suddenly in vogue, a steep trend poses problems for forecasting.
Having forecasting capability that identifies these items and understands their lifecycles early means you can ride the wave for as long as possible without being crushed beneath it once the hype eventually dies.
Once in a while a major event causes the tectonic plates of the economy to shift, and consumer behaviour undergoes a rapid change.
In such a situation, you must ask yourself: has this event created a new normal? Or is it a once-in-a-century anomaly, after which life will return to normal almost as quickly as it changed? How will March 2021 demand compare to March 2020 for grocery retail?
Having a forecasting solution which can establish new normal, but also get back to the old normal when applicable, is essential to forecasting correctly during the event, and in the aftermath.
If an external force triggers a stockpiling cascade it further complicates forecasting.
There are three levels of stockpiling: pantry filling, stock-up and hoarding. These typically translate to 1.5x, 2.5x, and >4x regular demand, respectively.
The first challenge when faced with an extraordinary event is to meet the initial stockpiling spike. This is critical for the business's reputation and revenue. To meet this challenge effectively, retailers require forecasting algorithms that quickly pick up on the unprecedented trend and adjust without prejudice to the new reality.
The second challenge is to alter this forecast once the external force subsides and/or consumer confidence is restored, to avoid overstocking the supply chain, binding capital and, eventually, creating waste.
The third challenge is to account for the event in next year's forecast, as the same period comparison might be different for what you normally see.
Of all the cases, odds are that the external force, and the rollercoaster it will entail, is where human intuition will fail us the most. It’s important to build forecasting systems that are efficient, and robust in normal, everyday business, but also flexible enough to pick up on new realities.
Well, none of these challenges are mutually exclusive: yes, new products are launched in trending segments during times of crisis. Having access to systematic, automated and robust analytics when working on advanced forecasting is crucial.
Today, there are technologies available to do better forecasts than ever before. Even during unprecedented events, modern machine learning algorithms, effectively orchestrated, can cope with multi-dimensional forecasting challenges surprisingly well.
Human intuition and experience should never be disregarded. They should be embraced, but supported by powerful technology.
Finally, to succeed in forecasting under the most difficult circumstances, retailers must be able to depend on:
A) Intuition and experience
B) Automatic data collection, that works with current systems
C) Machine learning models used at scale and short turnaround time of analysis
D) A user interface for decision makers, easy to use and understand
We're confident you're taking care of A. See how Formulate can help with B, C, and D.
Sign up to receive stats, data, and insights that will help you keep up with retail trends