How do you turn data into a 360° customer view?
Which technologies support data science initiatives?
How can detailed forecasts be made?
How do you turn data into a 360° customer view?
Which technologies support data science initiatives?
How can detailed forecasts be made?
Data science methods only provide usable information if the basis is correct. Inaccurate data always means inaccurate findings. Or, to put it another way: if the walls on the first floor are crooked, the building plans for the loft conversion are useless. If you carry on building anyway, you'll make things worse because you won't be able to keep up with the lamination in the end.
Energieversorgung Mittelrhein (evm) wanted to be able to track the customer journey in a more targeted way. Building on this, an automated and customer-specific approach based on contract and behavioral data was on the agenda. At the end of this value chain of data-driven insights, forecasts and trends were to be analyzed and used as economic decision-making aids. Back to house construction: in order to install the latest state-of-the-art solar technology on the roof, the right concrete first had to be mixed for the basement.
.png)
Together with taod, Energieversorgung Mittelrhein successfully completed the complex and time-consuming cleansing of its database. Using the record linkage method, the cleansed data was then compared to create a corrected history. The creation of a new cloud-based modern data warehouse in Azure also forms a reliable and high-performance basis for all future data science initiatives.
In a data thinking workshop and a subsequent data discovery workshop, evm and taod record the current situation. The identification of initial use cases ensures a detailed requirements analysis, which ultimately leads to an MVP approach. Sprint planning is carried out on the basis of the existing data strategy, which allows for permanent agile enhancements.
As the data in the billing system is of poor quality, cannot be checked across multiple attributes and similarity comparisons are not possible, the record linkage method is used. Based on the similarity estimates of the selected attributes, the record linkage model automatically learns how likely it is that two data records belong together. As these calculations require complex statistical procedures (such as expectation maximization), it is essential to set up an adequate cloud infrastructure.
Azure Synapse not only provides the necessary data connection and data preparation through ETL pipelines for setting up the cloud environment for Energieversorgung Mittelrhein, but also seamless integration into cluster compute instances in Apache Spark. With this tech stack, it was possible to automate existing approaches for unique customer identification, flexibly integrate the results into the new modern data warehouse and incorporate them into the business processes with adequate monitoring.
Together with taod, evm has successfully completed the complex and time-consuming cleansing of its database using the record linkage method. The creation of a new cloud-based modern data warehouse in Azure also forms a reliable and high-performance basis for all future data science initiatives.
Sarah Burdenksi
Team Leader Customer Loyalty & Analytical Marketing / evm
Make predictions, automate your processes and uncover trends and patterns with our AI consulting.
With Microsoft Azure Services and our Azure Consulting, you can extend the limits of your current platform.
How to recognize, promote and accelerate the potential of data initiatives for data science solutions.
