Best Practices Summary:

• A centralized data platform can help democratize access to data, improve data quality, and enhance understanding of data location and meaning for greater efficiency and effectiveness of delivering analytics. 

• Driving mentoring and learning for data scientists on both data and modeling skills as well as business acumen helps to drive analytics’ fit for purpose and adoption by the business. 

Case Study:

A large public utility in the Pacific Northwest established a data program to improve business outcomes. Example outcomes included predicting electrical outages before they occur, addressing customer needs before they called the call center, and segmenting customers to communicate with them optimally. However, the immaturity of data management and advanced analytics were roadblocks to meeting these objectives. Challenges included limited access to data across the organization, poor data quality, and lack of advanced analytic talent. 

The organization sought to address these challenges and opportunities through investments in analytics, building of a data foundation, and changing of the culture – all leading to greater adoption of data analytics. 

The first challenge was to focus on data quality of customer information.  A large effort to standardize and clean up customer data was prioritized. Data was profiled and cleansed while analysis was conducted to identify and implement better collection processes so that data would be cleaner as it was acquired. 

To address the data accessibility challenges with customer data residing in disparate systems, a centralized analytics platform called the “Platform of Insights” was developed.  The SAP HANA based platform brought together large data sets from both on premises and cloud data sources using virtualization.  The data platform democratized access to data initially for customer specific use cases giving the company a 360-degree view of their customers.  The platform has been expanding as more complex use cases require new information like utility asset data or meter interval data. 

It was also important to shift the organizational culture toward stewarding and using data along with the aforementioned improvements in skills and technology. To ensure quality and trust in the data, a data governance program was implemented that was steered by a data governance council. Data steward roles were designed and assigned. The data stewards profiled data for quality and made measurable improvements. Challenges with the understanding of location and meaning of the data were tackled by implementing a data dictionary. 

These shifts required a new data operating model that made accountabilities and responsibilities to drive value-added activities in the business clear. The operating model considered challenges such as ability to pay appropriate salaries to data science-skilled resources. The lines of business were lacking a salary class that would make hiring data scientists feasible, so the IT team worked with HR to create a new salary class that made hiring data scientists possible. The data scientists were assembled in a centralized group within IT. They entered both mentoring and learning programs that helped with modeling quality supported by peer reviews. An emphasis was placed on combining data skills and business acumen to help enhance the success of analytics adoption by the business. 

The improvements in the analytics, data platform, and culture of data stewardship helped to drive meaningful improvements in business operations. In the customer area for example, customer segmentation developed via analytics on accessible and higher quality data allowed the targeting of customers in collections who historically failed to pay but could now receive assistance from available assistance programs. The targeting and messaging activity drove more customers to adopt these programs lifting revenue from collections. As another example, electricity outages could often negatively impact customer satisfaction, especially when they were unplanned and/or not communicated in advance. With improved data management and stewardship to deliver quality data, advanced analytics could be used to more accurately predict the likelihood of outages based on inputs from utility grid assets that indicated overloaded transformers, often a precursor to an outage. With this information, planning teams can schedule maintenance and other activities to reduce the frequency of outages and improve the customer experience.