DIRECTV is a leading video and entertainment provider, using data across customer engagement, content consumption, subscriptions and billing to power decisions across marketing, product and operations. To unify this data, the team built Newton, a centralized dataset with over 1,200 customer-level attributes. With Databricks AI/BI, Genie and Databricks Apps, DIRECTV enables teams to ask questions in plain language.
What challenges did DIRECTV face with data and analytics?
DIRECTV’s data ecosystem spans nearly every aspect of its business, from customer behavior and content engagement to subscriptions, billing and product usage. This data is essential for evaluating campaign performance, understanding customer trends and guiding executive decision-making.
To bring consistency to this data, DIRECTV built Newton, a unified view of each customer with more than 1,200 attributes. Updated daily with historical snapshots, Newton became the foundation for analytics, machine learning and campaign targeting across the organization.
Marketing teams use it to measure campaign lift through control and test groups. Machine learning models rely on it for training and scoring. Executives use it to analyze trends across time, accounting for seasonality and major events.
But centralizing the data did not make it easy to use. “We had the same datasets built five different ways by five different teams. There was no single version of truth,” said Malav Shah, Director, Data Science at DIRECTV.
With thousands of columns and billions of records, analysts needed significant time to understand the schema, identify relevant features and write efficient queries. New users often spent weeks ramping up. Business users without SQL expertise depended entirely on technical teams to access insights.
Analysts relied heavily on ad hoc SQL, which introduced inefficiencies and strained shared infrastructure. Simple exploratory questions could take days or weeks to answer due to back-and-forth iteration.
DIRECTV needed a way to make Newton accessible to both technical and non-technical users, without sacrificing performance, governance or trust.
How did DIRECTV enable self-service, conversational analytics with Databricks?
To make Newton accessible, DIRECTV adopted Genie as the primary interface to its data.
Instead of navigating schemas or writing SQL, users can now ask questions in plain language and receive answers instantly. Genie translates those questions into optimized queries, retrieves relevant data, and returns results as tables, charts and natural language summaries. Users can refine results through follow-up questions, enabling rapid iteration and deeper exploration.
“At the core, we want users just to ask the question, and Genie figures out the right way to answer it,” Malav said.
Genie is effective because it is tuned to DIRECTV’s data model. The team incorporated example queries based on real analyst workflows, allowing Genie to accurately map business questions to the correct logic across a wide and complex dataset. This level of control was critical for ensuring accuracy.
Before adopting Genie, DIRECTV evaluated other approaches. DIRECTV selected Genie because it allowed the team to incorporate domain-specific guidance and example queries tailored to Newton.
To deliver this experience across the business, DIRECTV built a set of applications using Databricks Apps. These apps package Genie spaces into interfaces tailored to specific workflows, such as sales insights and customer analysis. Business users access these applications through Genie, a single-entry point for AI/BI Dashboards, applications and AI-driven insights.
Behind the scenes, data is governed through Unity Catalog, ensuring consistency and secure access across teams.
To support these applications at scale, DIRECTV uses Lakebase to store conversation history and enable fast, responsive interactions. This allows users to revisit prior analyses, iterate quickly and interact with data in real time. As usage scales, Lakebase supports the low-latency, high-concurrency access required for production AI applications.
To build trust in AI-generated outputs, DIRECTV implemented an evaluation framework using MLflow that scores responses based on correctness, groundedness and clarity. Each result includes a confidence score, helping users assess reliability before making decisions.
“We generate an evaluation score so our team can gauge reliability before acting on results,” said Santheswara Talluri, Principal Data Engineer at DIRECTV.
The system continuously improves by learning from real user interactions. By analyzing how users ask questions in production, the team refines prompts and improves accuracy over time without manual rewriting.
What business impact has DIRECTV achieved with Genie?
The impact has been noticeable across both technical and business teams. Tasks that previously required multiple iterations and deep knowledge of Newton can now be completed more efficiently.
"Before, it might have taken longer to get to the final answer. Now it's a shorter process," Santheswara said.
Analysts spend less time writing and optimizing queries and more time interpreting results and exploring new ideas. Business users who previously depended on technical teams to access Newton can now interact with it directly in plain language. Use cases are expanding across the organization:
Marketing teams build more precise audience segments and target campaigns using customer behavior and engagement data, enabling more relevant and effective outreach.
Product and analytics teams track engagement trends and build dashboards for leadership.
Executives explore historical and seasonal patterns to inform strategic decisions.
Adoption is growing. Today, hundreds of analysts use Genie-powered apps daily. Individual business units are also standing up specialized Genie spaces on their own datasets.
Beyond speed, the most significant change is in how teams access data. Users can explore ideas in real time and ask follow-up questions to refine their analysis.
Previously, Newton was used almost exclusively by technical teams serving business requests. Now, business users interact with the same data directly. "It gives more of our team direct access to the data," Santheswara said.
DIRECTV is continuing to expand this approach — bringing together AI/BI Dashboards, applications and Genie-powered insights in a single place.
* DIRECTV is a Databricks customer. This story was produced in collaboration with Databricks.