How a utility giant is using data analytics, machine learning to benefit customers

Utility giant EDF UK wanted to find a way to exploit its disparate treasure troves of data assets and create pioneering services for its customers using up-to-date data analytics and machine learning technologies. The answer to this difficult challenge lay in using less tech, not more.

Alex Read, senior manager of data platforms at EDF UK, says the company has embraced digital transformation during the past 12 months, moving from a disparate collection of bespoke and off-the-shelf systems to a tight enterprise data strategy based on the tactical use of cloud-based services.

“The less tech, the better — understand exactly the minimal amount of technology you need to arrive at the outcome you desire,” he says. “Previously, we had a vast technology estate that was borderline unmanageable. We now have a few technology components that just make our lives 10 times easier.” 

Read says this transformation process has encompassed two key strands: overhauling technology systems, and delivering business benefits. The experiences his team has been through during this process provide important lessons for all data leaders.

Transforming the technology platform

After 12 months of hard work, EDF UK now boasts a one-stop-shop for customer analytics that provides a clear understanding of the challenges the utility firm faces and the kinds of products its customers need.

Read says the integrated data platform — which uses Snowflake technology, AWS cloud services and other tactical solutions — is a long way from the provision the company’s staff had to rely on before. 

“We had a bunch of vendors, a bunch of data solutions, both on premise and in the cloud,” he says. “We had no central source of truth and no consolidated view of all of our data that we wanted to share across the business.”

Read says EDF UK relied historically on a plethora of bespoke and off-the-shelf tools. With no centralized platform, the data team struggled to meet the business’s demand for new analytical services. As a result, they often created siloed environments to support each niche use case.

Data science was particularly challenging. Despite two and a half years of effort to stand up a machine learning platform, the business did not have a scalable production environment. Key stakeholders at EDF UK were concerned the business wasn’t seeing a strong return on investment, so Read’s team focused on a new approach.

Information from across the business is now collected in an AWS S3 data lake and consolidated in the Snowflake Data Cloud. EDF UK then uses the Snowpark development framework to allow its data scientists to use Python and bring machine learning models into production on AWS SageMaker. 

The company selected Snowflake after an RFP exercise. It valued the platform’s cloud-agnostic, scalable and developer-friendly nature. About 90% of EDF UK’s legacy systems are now migrated to Snowflake. Read envisages Snowflake will be the sole enterprise data platform by the end of the first quarter 2023.

Moving disparate data from legacy systems to the cloud was a complex process. This “lift and shift” process was made easier by undertaking an optimization phase, where EDF UK worked with Snowflake to perform regular reviews that monitored the effectiveness of the integration. 

As well as Snowflake and AWS technology, EDF UK uses tactical solutions, such as Matillion for data extraction and transformation, Colibra for data catalog and governance and Apache Airflow for orchestration. Read says the result is a tightly integrated enterprise platform that gives users easy access to trusted sources.

“Having data in one central place means that whatever your role is — whether you’re a reporting analyst, data scientist or regulatory report developer — you have one central place where you can interact with data. That has been an absolute game-changer for us,” Read says.

Delivering business benefits through machine learning and data analytics

Read says cloud-based transformation will help EDF UK to compete with nimble startups that are disrupting the utility sector. Unencumbered by legacy ways of working, these lean organizations respond quickly to changing customer demands. By embracing digitization, EDF UK can respond quickly too.

“We believe one of the key differentiators we could have over some of our startup competitors is data,” he says. “We believe that we can make the most of our data due to the significant experience we hold and the platforms we use.”

One big benefit of EDF UK’s digital transformation is the ability to develop new machine learning products quickly. The firm is now using AWS SageMaker and Snowpark to create and deploy products in weeks rather than months or even years. 

Read says his team is producing four times the number of data science products it was historically. They’ve developed models to identify key business challenges and are creating data-led products for customers, boosting satisfaction and retention.

“We want to give our employees a better idea of the material impact their actions have on each customer,” says Read. “That creates massive insight for us as a business.”

EDF UK is using its models to analyze which customers are most likely to purchase an electric vehicle. The firm is also using its data and machine learning models to identify customers who might be vulnerable to financial difficulties.

With this richer insight, EDF UK is developing products that not only reduce business risk but that support customers. The data team has worked with the UK’s National Grid to build a “winter turndown” tool, which shows how customers can cut energy use at peak times and potentially reduce bills. 

Energy Hub is another product from the data team. The analytics platform, which is available on the web and in an app, monitors energy use and makes predictions, so customers can take control of their energy use.

Read’s team continues to look for new opportunities to exploit data. They’re currently thinking about how to create models that ensure the company buys its energy at the best price in an increasingly volatile market due to Russia’s invasion of Ukraine. 

The aim going forward is to ensure the business can self-serve. With the platform and models for data analytics in place, the onus is now on the business to develop use cases. So, while the past 12 months have been about working hard to build a platform for the company’s data-led endeavors, the hard graft is only just beginning. 

“The next step of our data journey is putting data back into the business, so that their use cases can be served,” says Read.