Point of Sale Assortment Management Using Machine Learning

By combining predictive analytics models and parameters that are directly or indirectly related to product profitability, items at the level of an individual location can be grouped based on profitability, that is, not based on past events, but future trends. For retailers looking for ways to increase revenue while reducing costs, predictive analytics can offer some intriguing possibilities

Creation of assortment lists
Many companies are listing new products and creating new promotional ideas to increase profits in their companies. Artificial intelligence (AI) is a powerful tool for determining the profitability of individual products. Product assortment people usually choose which items to include in their products based on business goals. Choosing profitable items helps businesses achieve sales success. Corporate buyers and suppliers choose which items to list based on current customer requirements. For example, a restaurant owner can create a new dish by choosing ingredients to buy from their suppliers. However, some vendors have a large selection of items they can sell – maximizing profits for the business owner. A supplier may offer a wide variety of options for a specific item, such as dinnerware or restaurant tablecloths. In this way, the business owner can adjust the list of items to his needs. In addition, it is an easy way for him to maintain or increase profitability.

When determining which items to offer, companies should consider product demand as well as the difficulty of sourcing and manufacturing each item. For example, some consumer goods are popular among consumers because they are cheap and easy to use. Additionally, many such items are available through multiple different factories and suppliers, reducing the risk of creating a successful product line. On the other hand, unique items may be more difficult to produce due to high production costs or licensing requirements from copyright holders. Determining which items to sell is a process of extrapolation that requires good business knowledge and financial acumen. Business owners typically choose profitable items when creating new offerings or changing the assortment for sale at individual locations.

It used to be thought that assortments with a wide selection of items were the most profitable because they most closely corresponded to the current needs of customers. However, by combining predictive analytics models and parameters that are directly or indirectly related to product profitability, items at the level of an individual location can be grouped based on profitability, that is, not based on past events, but future trends. Based on this data, the procurement of items can be managed and the profitability of each location can be directly affected. These calculations are done weekly to manage profitability over the same time interval. Based on this, mathematically managed assortment lists are created with a focus on the profitability of individual items at the location. From these lists, it is possible to create global assortment lists at the level of individual regions or types of sales locations. Of course, we are talking about mathematical and not marketing assortment lists that include some other relationships. For retailers looking for ways to increase revenue while reducing costs, predictive analytics can offer some intriguing possibilities.

Calculating Item Profitability Ratios at the Location Level
Profitability is typically tracked “backwards” based on sales, however our approach is to use item profitability as a parameter for future ordering. In order to be able to use this, we calculate the profitability coefficient, which provides us with greater or lesser “bravery” of the model when ordering. The important thing is that the coefficient is calculated at the level of each location and then enables different behavior when ordering items.

The process of calculating the profitability coefficient is carried out weekly, then the items are classified into ten profit categories at each location and assigned a profitability rating for the next week, and based on this data, corrections are then made in the ordering of items.
How an example of different ordering based on different profitability coefficients looks like can be seen in diagram 1.

Calculating item profitability ratios at the location level is critical in business decisions regarding promotion strategies, production, and inventory management. By understanding how various factors affect the sales performance of your product, you can make smarter business decisions regarding product quality, advertising and packaging. Ultimately, this will help you increase your profits with smart ways to determine the optimal number of items on the shelf at each location.

Determining the optimal number of items on the shelf (eng. facing)
The marketing strategy can be optimized based on the results of the machine learning model. For example, if a machine learning model determines that an item should occupy more than one location—such as multiple shelves in a retail store—this must be accounted for during implementation. On the other hand, if there is only one location for this product, planning must ensure that each customer sees at least one copy of this item during their visit to keep sales high. The number of items on the shelf is conditioned by different decisions and relationships, but by using predictive analytics and the profitability coefficient, it is possible to calculate the optimal number of pieces of an individual item that will be displayed, as well as the optimal position of the item on the shelf.

The machine learning approach to assigning an item to a category on a shelf is complex, considering the attributes of each item before deciding where to place it. There must also be enough space on the shelves for all products so that no customer is inconvenienced when purchasing items in the store. Planning must also take into account how products move between different locations so that sales remain high in all locations where those products are sold.

To perform this analysis, data from a predictive sales model is used, from which a profitability rating is calculated and combined with current data on the number of items on the shelf. A comparative analysis is then carried out to identify items whose number on the shelf deviates significantly from the forecast. Based on this and the profitability assessment, recommendations for changes and repositioning are given. After identifying these locations, analysts can use their knowledge of consumer buying habits to determine which units should be placed at each location on the shelf. Likewise, the sales growth potential of individual items can be analyzed and changes in positioning can be proposed. Once these items are on the shelves, they should get the best exposure and sales thanks to better visualization by consumers.

The basis of the process is the sales prediction for a particular period, which is always increased by a minimum amount. The minimum quantity is the number of items that always “must” remain on the shelf so that customers do not have the impression that the store is empty. However, it is important to emphasize that when ordering, we take into account the minimum quantity that can be consumed, and it is not seen as locked goods on the shelf, but goods that are also available for customers, but which ensure that the shelf does not remain empty. The usual process is to take the initial values ​​(of course if they exist), in order to know which shares ensure that the appearance of the store does not give the impression that it is “on inventory or closing”. Then we start with the dynamic determination of these quantities and corrections, and the result is dynamic limits of minimum quantities that then vary depending on the season and promotions.

For more information, contact us at: www.aduroidea.com

Prethodni članakFortenova u prvih devet mjeseci s prihodima iznad 30 mlrd kn
Sljedeći članakGet ready for 25 days of giveaways at Ja TRGOVAC internet portal!