New Item Sales Forecasting Using Similar Store and Item Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retailers face challenges in optimizing store assortment due to the lack of historical sales data for new items, leading to inaccurate demand forecasting and inefficient product selection, especially when transitioning from e-commerce to physical stores.
Innovation Solution
A system utilizing a feed forward deep neural network with a hierarchical architecture to forecast sales data of new items by leveraging item, store, and sales features, even with sparse input data, to provide accurate demand predictions for assortment optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing forecasting methods are used for new items without historical sales data, then the forecasting process can be executed, but the prediction accuracy deteriorates due to data sparsity resulting in all zero values
Solution Approach 1:
The patent introduces an intermediary mechanism by leveraging historical sales data from similar items and similar stores as a bridge to forecast sales for new items without direct history. The system uses item similarity metrics and store similarity metrics to find comparable entities, then applies their sales patterns as intermediate evidence to predict new item performance, avoiding the zero-value problem
Solution Approach 2:
The system performs preliminary actions by pre-computing and storing historical sales data from similar items and similar stores before the forecasting need arises. This pre-prepared data foundation enables rapid and accurate forecasting when new items are introduced, eliminating the need to start from scratch with zero historical data
2Device complexity
If traditional linear relationship methods are used, then the model complexity remains low, but the ability to handle huge amounts of data and capture complex patterns deteriorates
Solution Approach 1:
The patent transitions from linear relationship modeling to multi-dimensional analysis by incorporating item dimensions (category, brand, price) and store dimensions (location, size, customer demographics) simultaneously. This dimensional expansion enables the system to handle huge amounts of diverse data and capture complex interrelationships that linear models cannot detect
3Adaptability or versatility
If retailers refresh assortment frequently to meet evolving e-commerce trends, then customer preferences are better matched, but the time and resources required for assortment optimization increase
Solution Approach 1:
The forecasting system operates as a self-service tool that automatically generates sales predictions for new items using pre-configured similarity metrics and historical data. Retailers can independently evaluate potential assortment additions without requiring extensive manual analysis or external consulting, enabling rapid iteration and frequent assortment refreshes with minimal time investment
Data Source
AI summary
Systems and methods for forecasting sales data of items that are new or missing historical sales data at a physical retailer store are disclosed. In some embodiments, a disclosed method includes: receiving, from a computing device, a forecast request seeking sales data of an item if the item is offered for sale at a physical store in a future time period, wherein historical sales data of the item at the physical store is not available; determining, based on the forecast request, at least one relevant feature related to the item or the physical store; computing, based on a machine learning model and the at least one relevant feature, forecasted sales data of the item at the physical store in the future time period; and transmitting the forecasted sales data to the computing device.