About

The Client is a European dairy manufacturer that is holding leading positions on the domestic market, as well as exporting their wide portfolio of goods to 20+ countries worldwide.

Challenge

The Client was dissatisfied with unstable sales performance vs. the plan and, consequently, missed sales targets. They wanted to get accurate sales forecasting and achievable targets per product category, per brand and per store. Besides, the Client wanted to understand the gap between their plan vs. fact sales performance and identify potential for improvement.

Solution

VolgoTechnologies data scientists started with cleaning the historical data. For example, our team removed duplicates by narrowing down differently spelled street names in the stores addresses to one correct record.

Our data scientists had a detailed look at plans vs. actual sales per product category, per brand, per store and per region. To ensure accurate sales forecasting, the team excluded the influence of promotions and developed an algorithm that allowed selecting the most relevant statistical model. Depending on how far back the sales history went, one of the four models was automatically chosen: linear regression, autoregressive integrated moving average (ARIMA) model, median forecasting or zero forecasting.

Staging

Datawarehouse

Dataware House

Desktop Application

Results

The Client received an accurate sales forecast built on statistical models and algorithms applied to historical sales data with a growth rate added.

Technologies and Tools

Microsoft SQL Server (a data warehouse), Microsoft SQL Server Integration Services, Microsoft SQL Server Analysis Services (online analytical processing), Python, Microsoft Excel, Microsoft Excel Power Pivot.