In-store performance prediction refers to sales forecasting for products in brick-and-mortar stores. While it is one of the most important skills a retailer needs, it is also one of the most difficult, combining retail data with marketing, geospatial, weather, and other data.
A good model will enable stores to reach success on the most basic and important measure: sales. Additionally, a good model will reduce the risks of both under-stocking and over-stocking inventory.
The most important internal data for a good in-store performance prediction model is historical sales data. In particular, sales data viewable down to the weekday and correlated with promotions and holidays. Website visit data enriches the internal data for this model, as well.
Essential external data includes the holidays and events like preparations for children going back to school. Additional data includes weather as that impacts footfall and the general economic situation of the region the store is in.
Additional external data that may enrich an in-store performance prediction model may include test markets. Similar to focus groups or surveys, potential customers give their opinions on upcoming products so companies can determine the likely best sellers.
Some challenges to building a strong performance model affect all industries. Trends, for example, and changes to trade or tariffs. And, of course, economic changes affecting a region or population.
However, some industries face additional struggles—in particular, the fashion and electronics industries. These fields are particularly sensitive to trends, have short sales lifecycles, and have very little historical data to rely on as people often throw out the old with every season or new invention.
“As online buyer behavior remains strong, we expect offline sales to decline by 6.6% in 2020,” Forrester reports. [However,] bullish Deloitte predicts e-commerce holiday retail sales to grow between 25% to 35% from November through January, reaching $182 billion to $196 billion in total.