The ability to predict how much a food item will be consumed on a specific day is usually something an experienced store manager does in today’s world. But with Baidu’s latest experiment using a machine learning algorithm, this job may well become obsolete.
With historical data from 70 metrics including store food purchases, sales, the weather and festivals, Beijing-based Baidu said it has developed a model that can predict store sales for the next day as a reference for the store manager to decide on the quantity of food products, such as rice boxes and sandwiches, needed to meet the customer demands without creating excess waste.
Using data from over 70 metrics, Baidu has developed a ML model to predict the sales in the store for the following day. The metrics that are being tracked include food purchases, sales, weather, and festivals, among others. The model achieves a two-fold use for the store manager:
- it helps the manager decide how many items need replenishing,
- it helps eliminate, or reduce, wastage of the products
This experiment was conducted over a period of 10 days at 10 convenience stores in China. According to Baidu, the model helped increase the average profit across these store by 20% and cut down on the waste by 30%. Mr. Liu Yongfeng, the senior project manager of Baidu’s deep learning platform, revealed that they are planning to embed this technology across 200 more stores in China’s Wuhan city this year.
Our take on this
This is a slightly different take in the retail space – rather than focusing on the customer, Baidu are intent on solving an industry level problem. Amazon, Bingo Box, etc. have opened stores that help the customer avoid queues using facial and voice recognition. Baidu, on the other hand, are focusing on the back end. Fresh food, if not consumed within a stipulated amount of time, has to be thrown away. Using their model, the company is aiming to reduce the excess waste and improve the bottom line of each store.
Another issue in this field is the high turnover of store managers. Quite often, they leave without passing on their entire knowledge to the successor. This technology will help even a not-so-experienced person take over and ensure the stores are replenished with the optimum quantity.
You can also read this article on our Mobile APP