For Electrolux, understanding customer purchasing trends is key to accurately predicting sales forecasts and revenue projections. But with a highly complex global infrastructure involving 400+ distribution centers across 155 markets that move over 60+ million products per year, doing so is easier said than done. With Databricks, Electrolux has been able to leverage machine learning to improve sales forecasting and streamline the supply chain, allowing them to predict revenue and improve customer satisfaction by delivering products in a timely manner.
It’s impossible to predict the future. But with the proper level of insights, it is possible to come close. In the case of Electrolux, they set forth to better understand what makes a customer tick with the goal of improving marketing campaign ROI and sales forecasting.
Electrolux has a highly complex global supply chain that boasts 400+ warehouses and distribution centers moving over 60+ million products per year. With this complexity comes a number of challenges that can span competitive pressures, volatile materials costs, and more. “Rather than boiling the ocean, our data team sought to focus on leveraging data science to improve the sales and marketing parts of the business,” explained Johan Vallin, Global Head of Data Science at Electrolux.
Historically, the team would take an Excel spreadsheet, enter in historical information (i.e., pricing, trade shows, online ratings, reviews) stored in multiple data sources to make strategic decisions. This process was not only highly resource-intensive but resulted in inaccurate forecasts without a clear picture of the influence of marketing campaigns.
“Because of the complexity of our supply chain and the number of models to manage, we needed a better way to not only manage our data assets, but also version control all of the different models,” said Johan. “We now handle all of this through the Databricks platform.”
With Databricks, Electrolux can now easily ingest massive volumes of consumer and product data and leverage machine learning at scale to better understand how to optimize the value chain through improved marketing ROI and sales forecasting.
When dealing with multiple data types and sources, data reliability is key for downstream machine learning. With Delta Lake, the data team at Electolux is able to easily ingest millions of data points and build robust production data pipelines at scale without fear of data quality issues.
These pipelines then feed into 1000s of models that are being constantly improved upon. MLflow streamlines the complete machine learning lifecycle for Electrolux. “With Databricks, we can now track different versions of experiments and simulations, package and share models across the organization; and deploy models quickly,” explained Johan. “As a result, we can iterate on predictive models at a much faster pace leading to more accurate forecasts.”
It is also critical for the data team to serve up actionable insights in a way that non-technical team members can leverage. Through its native integrations with the Microsoft cloud ecosystem, Databricks empowers Electrolux to develop a sales volume forecasting dashboard that provided insights — including price trends, competitive influences, campaign costs and activities, and more — to optimize sales and marketing campaigns. In all, the unification of data science, engineering, and the analyst team under the same platform has boosted cross-team collaboration, unlocking new data-driven opportunities to improve sales and operations across their business.
With Databricks at the core of Electrolux’s sales forecasting engine, the complexities of their global supply chain is now considered a prized asset and competitive differentiator as it feeds 1000s of models that are continually getting better and smarter.
The company is now able to make more accurate predictions on how consumers and the market may behave in the future. These newfound insights have improved all critical facets of their supply chain, particularly with the impact sales and marketing campaigns can have on sales forecasting. With Databricks, not only are sales promotions more on-point, but Electrolux’s ability to deliver the right products to the right consumers at the right time is directly benefiting the bottom line, and proving to be a trusted vehicle for sustained growth quarter after quarter.
Databricks allows us to adopt a more scientific approach to how we run our sales operation. As a result, we’ve tripled our forecast accuracy of sales promotions.”
– Johan Vallin, Global Head of Data Science, Electrolux
テクニカルトーク（Spark + AI Summit EU 2019 より）