Skip to content
Contact us
  • There are no suggestions because the search field is empty.

Store Transfer's Demand Forecast

The platform uses a decision-tree model trained on hundreds of millions of data points extracted from historical data, using 60+ defined variables it will arrange to come up with the best forecast outcome. 

Imagine you want to forecast the sales of a particular product—such as white pants during a busy seasonal event. To calculate the forecast, Nextail will complete the task respecting the following steps: 

  1. Establishing the Initial Prediction (tree base):
    The algorithm calculates a forecast using sales from the first two years using one of the variables as a starting point (for example, the product family). This initial forecast is then compared to the actual sales of the third year and a forecast error is calculated. At this point, the error is big since the system is only using one variable to determine the forecast. 
  2. Refinement Through Variables (branches generation and organization): To improve the accuracy, the algorithm incrementally adds more specific criteria (time of year, then store type, then geographic location) thus adding branches to the tree. At each branching decision, the objective is to reduce the forecast error from the previous calculation. For instance, it might introduce the month of July as an indicator since sales behavior varies with seasonality. By doing so, the forecast is adjusted to better reflect real-world outcomes.
  3. Optimization with an Ensemble of Trees (tree selection):
    Instead of relying on a single decision path (only one tree), the algorithm automatically develops multiple decision trees, experiments on combination, order and weight of variables and compares their predictions. By evaluating over 60 different variables in various sequences, the system can identify and select the configuration that yields the most precise results.

Nextail is using an advanced machine learning technique that builds a strong predictive model by combining multiple decision trees. Each new tree seeks for better forecast results. This iterative approach enhances both prediction speed and accuracy.


Store Transfer Methodology in a nutshell


When running a Store Transfer Scenario the system will calculate:

  • The demand forecast of each product where it’s available at that moment (origin store). This is key to assess what could be the sales generated if the item was not moved anywhere.
  • The demand forecast of each product in the rest of the stores (potential destination stores). This is key to assess what could be the sales generated in each of the stores where it would be possible to transfer the item.


Thanks to this methodology, the system now has all the information needed to be able to compare the outcome of all possible options and therefore determine which products and through which trips could the scenario generate the highest Sales Increase.

Thanks to the volume of data handled and the degree of complexity behind this approach and its calculations, Nextail is able to minimize excess stock, reduce lost sales opportunities and ensure that inventory is allocated where it could have the greatest impact for the business.