Why are historical sales not enough for quality forecasting and ordering?
Have you ever heard the claim that all you need for accurate forecasting is to just supply historical sales and artificial intelligence (AI) will take care of the rest? If so, you better watch out.
Yes, it is undeniable that analyzing historical sales and buying patterns is fundamental, and machine learning has significantly improved the performance of traditional algorithms in recent years, allowing for faster and more accurate estimates.
However, the data on historical sales and buying patterns alone is not enough to solely base an accurate forecast on.
While AI definitely brings significant benefits to the world of retail, unless the AI is provided with detailed data, its predictions will simply not be accurate enough. Let's look at a specific example showing why it is necessary to include more than just historical sales data.
Factors to consider for demand forecasting
Imagine the following scenario: AI analyzes a product that has not been selling for weeks, but has recently experienced a significant increase in sales and is stabilizing at this new, higher level.
What correct forecast of future sales do we expect AI to deliver?
The correct answer is not clear-cut because there are other factors besides historical sales that affect the forecast. Let's consider a few examples:
1. Insufficient inventory
- The volume of sales may remain at a higher level if a product was not sold in a previous period simply because it was out of stock in a particular store and now we are capable of replenishing it in sufficient quantities in the store again.
2. Sales promotions
- The volume of sales may drop back to its original level if the promotion that increased demand ends and price of the product is not expected to be at that level until the next sales promotion.
3. Seasonal promotions – yearly periodicity
- The volume of sales may fall, but it will return to high levels in a year's time, such as with products like school supplies that the parents buy in large volumes once the school year is about to begin.
4. Seasonal promotions – non-yearly periodicity
- The volume of sales may increase again, but not in a year's time if, for example, we are looking at a product that is typically sold before Easter.
Customer buying behavior is influenced by more than just promotions. There are many other factors that need to be taken into consideration that have a direct impact on sales performance. These include events such as football matches or concerts taking place near the store, a specific day of the week or whether the store is new to the market. Each of these aspects can have a major impact on customer decision making.
These examples clearly show that in order to achieve accurate forecasts, it is essential to include not only historical sales data, but also information about inventory and past and future events that may affect sales. AI will then be able to assess the impact of these factors on sales more accurately and offer more precise predictions.
Precise forecasting is key when it comes to automated ordering
Demand and order forecasts are closely intertwined. Thanks to high quality forecasting and integration with planograms, AI can select the optimal forecasting model and design orders with higher accuracy than the average store manager.
Especially in the first few months, investing time and effort into setting up everything properly, is worth the effort. By doing so, you can give the AI much better idea of what we want to achieve and what to avoid, resulting in better automated responses to the specific needs of your stores.
The adoption of automated, or at least semi-automated, ordering will ensure your stores have the right products in the right quantities and at the right time, leading to increased customer satisfaction and sales growth.
By choosing Quant Demand Planning, you can be sure that it will accurately forecast in all of the scenarios described above as long as you provide it with the necessary data.
Petr Kavánek
CEO / Co-owner | Quant Retail s.r.o.