Go back to list
Demand Management Replenishment

Is forecast generation using artificial intelligence for everyone?

Without historical sales data, it is impossible to generate demand and sales forecasts using artificial intelligence. So, in theory, there are industries where the power of AI and ML in the context of forecasting will not be used. And in practice? That is not true.

What can AI offer to companies which routinely lack sales history? Frequently rotating merchandise and regular new arrivals are a daily occurrence for fashion retailers. The cosmetic sector is another example of an industry that also faces similar challenges.

By definition, a new product has no sales history. In the fashion industry, about 90% of products are dedicated to a specific moment in the season. This means that one can never expect historical sales data older than a few months. What retailers have at their disposal, however, is sales data for other discontinued merchandise, which can be put into groups or categories and demand can be forecast based on them. Therefore, the power of AI and ML in industries where new products are sold on a daily basis manifests itself in generating forecasts based on reference products (i.e., similar goods: of the same cut, color, size, length, etc.), and products belonging to the same category.

For example: we want to forecast sales of T-shirts and optimally stock the stores with them. It will be difficult to generate demand forecasts for each model, color and size. However, it is possible to aggregate historical sales data for all T-shirts. We extract information on the cut, size, price, color, length, etc. Based on this, it is possible to draw accurate conclusions for our group of T-shirts with specific characteristics.

It is also worth taking replenishment into consideration. For replenishment, both forecasts and inventory stock must be examined. A platform that analyzes data based on AI and ML is able to make accurate cross-analysis and choose the right scenario. To illustrate: there are 100 SKUs of T-shirt A and 1,000 SKUs of T-shirt B. Knowing that all T-shirts should appear in store 10 and having determined the rules for generating forecasts beforehand, the platform will draw adequate conclusions. It will suggest providing 8 T-shirts B and 2 T-shirts A.

In this way, we make forecasts in relation to product groups, while also using artificial intelligence to create store demand lists and picking lists. Ultimately, we deliver the specified demand for a specific product to a specific store.

Would you like to know more?

Subscribe to our newsletter

Newsletters are sent in email format, no more often than once a month or immediately in case of significant news/changes/educational content. More in the terms and conditions.