Introduction
Replenishment objective
The Nextail replenishment module calculates a demand forecast for every SKU in every store and aims to maximize sales by placing each product in the stores with the highest probability of selling it. Our model, based on meritocracy, intends to send only the stock that the store is going to sell, keeping as much as possible for future replenishments. This avoids overstocking some stores and ensuring stock for others while being quickly reactive to changes in demand.
How do we do this?
A replenishment scenario is the result of combining our Demand Forecast, being this what we expect to sell, a probabilistic forecast at the SKU and store level that Nextail generates, with the Optimization, taking into account the stock positions and other factors like sales threshold or residual value, to decide how many units we need to send so that we fulfill the previous demand, plus some restrictions the user can add ad-hoc depending on business needs.
The algorithm in this solution is based on the following principles:
- Demand driven: Our Machine Learning Model generates a demand forecast for each sku-store, each day.
- Global optimization: The replenishment cycle is calculated across the entire network to ensure the stock is fairly optimized globally.
- Robustness over accuracy: Avoiding big mistakes takes precedence over increasing accuracy in most cases.
- Meritocracy: The starting point of the replenishment calculation is the existing stock, not store demand.
