)
Artificial Intelligence for Retail: AI-Driven Predictive Replenishment
Scenario
Working closely with our client, the need emerged to implement an innovative and efficient approach to the replenishment of shop merchandise, especially for low-turnover product categories (infrequent, sporadic and therefore not easily predictable sales).
The challenge was clear:
1. Avoid stock-outs, which would lead to lost sales
2. Reducing overstock, in order to make the occupation of warehouse space and resulting costs more efficient
3. Optimise the distribution of goods in different markets/channels
Solution
The sales forecasting solution conceived by Technology Reply places its core on Artificial Intelligence and the boundless resources of the cloud.
Starting from the basic idea of defining a “dynamic replenishment policy” to avoid stock-outs and at the same time trying to avoid unnecessary over-stocking, the predictive process is based on the ‘periodic review’ of the ‘reorder point’ (minimum quantity of stock, below which a replenishment order is triggered, to guarantee the availability of sufficient stock to meet demand during the lead time), correlating as many variables as possible.
The application takes as input the whole series of data necessary to formulate the forecast (sales, stock movements, promotions, weather conditions, footflow, delivery times... ) and, through its AI-engine (it exploits a Transformer model, the same technology behind ChatGPT), outputs the sales forecast necessary to evaluate a possible restocking.
For instance, by correlating weather conditions, it is possible to avoid re-stocking predominantly summer footwear types if the weather is particularly bad or vice versa.
We implemented our solution in the Oracle Cloud Infrastructure environment, thanks to the flexibility derived from the development language used (Python), the theoretically unlimited potential of the cloud and the customisation of an AI-based predictive engine that takes into account multiple variables, Technology Reply's solution provides the customer with a scalable and customisable forecasting and data analysis tool based on individual needs, allowing for much more accurate forecasts than the classic solutions available on the market.
Benefits
In this context of streamlining and optimising the replenishment process, the solution developed by Technology Reply has a number of key advantages for the customer:
In summary, the solution offers an effective forecasting and analysis tool (it supports the identification of significant correlations between data), because it combines the possibility of accurate forecasts with a broad customisation of the variables influencing replenishments.
Efficient re-assortment process obviously has its side effects on environmental sustainability, production efficiency (avoids production surplus), as well as reduction of costs related to transport, warehouse (stock) and marginality (e.g. sales).