585GOLD Jewelry store chain

GoodsForecast's cooperation with the 585GOLD jewelry store chain began back in 2016 with a pilot project to optimize product mix planning in the chain's stores. The first phase of the project involved research work aimed to develop mathematical models for handling illiquid goods and further issue recommendations to the customer on how to move goods between stores and remelt products.

Work stages:

1. Analyzing and understanding data.
   • Uploading data and primary analysis of whether they are consistent with business requirements, whether fields in tables are correctly filled in, relationships between entities as well as the volume of data of each type.
   • Analyzing sales history, product and store characteristics.
   • Identifying the main statistical characteristics of sales and the main factors influencing demand (promotion campaigns, price changes, width of the product mix, potential additional factors).
   • General report.
2. Creating a basic sales forecasting model..
   • Implementing a mechanism for launching numerous experiments to analyze the quality of prediction models and analyze the results.
   • Experimental work on model quality assessment and parameter selection. Modifying existing forecasting models or creating new ones based on them (optional).
   • General report.
3. Creating a price sensitive sales forecasting model.
   • Additional analytical tools for analyzing sales dependence on prices in detail considering various factors.
   • Creating a hypothesis and building a model of demand dependence on price.
   • Iterative work to conduct experiments, analyze results and modify the model.
4. Implementing a mechanism for issuing recommendations on handling illiquid inventories.
   • Implementing calculations to check if remelting products considering the costs involved is profitable.
   • Implementing an algorithm that evaluates the cost-effectiveness of moving goods to another store, price reduction and product remelting and selects the best option as the outcome.
   • Analyzing algorithm performance based on fresh data received from the customer, creating a pool of recommendations.
   • Adjusting models considering comments issued by the customer.

A corporate training was organized for the representatives of 585GOLD as part of joint work on the product mix planning optimization project. Specialists of the customer’s analytical department attended a course on machine learning organized at the Computing Center of the Russian Academy of Sciences.