George Santayana, the Spanish-American philosopher, famously mused, "Those who cannot remember the past are condemned to repeat it”, but Mr. Santayana didn’t foresee the Era of Big Data.
Traditional demand analytics analyzed historical data to spot trends that would likely repeat, but today’s demand forecasting is different and by using real-time data and advanced modeling, it produces accurate, relevant forecasts.
Market changes are both sudden and significant, so this couldn’t have come at a better time, and in this post, we’ll discuss the importance of understanding true demand, while also examining some of the opportunities and challenges for demand forecasting today.
True demand and the opportunities of demand forecasting
Imagine, a clothing store sells out of a pair of jeans, just two weeks into the month, all because of supply chain delays and inaccurate forecasting.
How many more pairs of jeans would the company have sold if they had them in stock? Had they’d added another shipment, would the store have had enough? Or, would another full shipment have been too many?
True demand means the exact demand for a certain product and understanding true demand is critical for both e-commerce and brick-and-mortar stores. Too much of a product means the store will have to mark down prices to get rid of it, while too little product results in lost profits.
Demand forecasting brings us closer to knowing true demand
Obviously, understanding true demand is easier said than done. Aaron Hoffer, Alloy’s lead data scientist, says that forecasting sales, rather than store demand, is a common demand forecasting pitfall.
Out-of-stocks suppress demand because they don’t account for potential lost sales. Going back to the aforementioned store example, does it make sense to use the sales number of an item that sold out after two weeks, when predicting future sales?
Alternatively, demand forecasting uses real-time sales and customer data to account for factors like out-of-stocks.
Demand forecasting allows us to drill down to granular levels
When companies focus on the total sales dollars, store-level differences are glossed over. For example, a chain with 20 stores in a certain region wants to predict sales dollars at the regional level, and then proportionally push those amounts to the stores.
Let’s say high sales at 15 of the stores drove the forecast. Would that result in an inventory overstock at the 5 stores with lower sales?
Demand forecasting negates the argument that store-by-store analysis is too time-consuming because it uses algorithms (e.g. the multi-arm bandit algorithm) to generate best-case scenarios that across all stores.
Despite these opportunities, there are still many challenges for demand forecasting
We need to understand how to use big data
Still think “big data” is a buzzword? Troves of enterprise data hold key insights that can increase productivity, spot trends in customer demand and reduce operational costs.
But, big data can be as much a burden as it is an asset and too often, users get lost in their data, and thus unable to find the necessities for analysis.
We need to adapt our workflows
The previous point speaks to a wider issue in new demand forecasting methods, and it’s best to incorporate them into your operations.
These new tools will affect departments ranging from marketing and sales, to data science to supply chain management. It’s common for these departments to stick to their old workflows, and then try to fit demand analytics insights in afterwards.
But, what’s the use of real-time sales data if your sales team only reviews performance on a weekly basis?
Demand forecasting is ever-evolving
Demand forecasting is changing e-commerce, consumer goods and retail, and companies can use real-time data for more reliable forecasts, ones that don’t fit with historical trends.
These changes are just the beginning, and the development of IoT and advancements in artificial intelligence and machine learning will only advance forecasting.
Organizations, regardless of size, need to leverage demand forecasting to remain competitive. Those that do will continue to satisfy customer demand and reduce their own lost profits and markdowns, while those who don’t will likely struggle to compete against faster delivery times and better-stocked shelves.