Following several customer requests, Webmarketing Booster wanted to add a sales prediction function to be integrated into its global solution, which had to take into account many and varied internal and external parameters. Thus, to provide a prediction, we designed a predictive algorithm that could be integrated into their platform.
Neovision created a technology to predict the sales of e-commerce sites. To do this, a state of the art of the different algorithms and forecasting methods was developed. Neovision then selected and used neural networks for learning and prediction.
Webmarketing Booster can now offer its customers a predictive function, based both on internal data, but also on external parameters strongly correlated to the variable to be predicted (weather, vacations, sales…). That way, end users can now anticipate sales volume and better meet demand while optimizing inventory management.
In the early days of Neovision, the company was based at Tarmac, Inovallée’s startup incubator, which was also the case for the company Webmarketing Booster, publisher of software solutions for e-merchants.
Not only did the employees get to know and sympathize with each other, the two companies also quickly understood that they could collaborate. Indeed, Webmarketing Booster designs, develops and markets software solutions to improve the performance of e-commerce sites.
Faced with this observation, the use of artificial intelligence seemed more than pragmatic. Even more so when several of Webmarketing Booster’s clients brought up a very specific need: to predict daily sales volume. To make this project a reality, we were able to count on a client of WMB (Webmarketing Booster), a major German shoe manufacturer, and exploit its data.
WMB asked Neovision and its expertise in predictive analytics to design and develop a sales prediction solution.
Before starting the project, Neovision discussed with WMB to gain access to its client’s sales history and market expertise. To be able to predict events, you need to know the factors that influence them.
As a result of these talks, Neovision decided which would be used, whether internal (sales history, sales per item) or external (temperature, precipitation, vacations, sales period, Google trends). This data was then exploited by the neural network for learning and prediction.
With this achievement, Neovision enabled WMB to make a sales prediction solution available to its customers. With this technology, e-merchants were able to improve their market visibility and reduce their storage costs. Previously, with its traditional prediction method, WMB’s customers reported an average 25% gap between forecasts and reality. With Neovision’s predictive solution, that gap has been reduced to 15%. The result is efficient inventory management (less overstocking, less out-of-stocks, etc.).
PREDICTIVE ANALYSIS, MACHINE LEARNING
16 October 2020
Analyse Prédictive, Machine learning