Comcast Advertising allows brands to connect with nearly 125 million U.S. households through multiscreen TV campaigns that span traditional and streaming platforms across all 210 designated market areas. With that scale comes massive volumes of data and the need for real-time insights. In the past, data scientists struggled to translate models into business-ready products. With Databricks Apps, data scientists have unlocked interactive forecasting dashboards that accelerate decision-making and enhance campaign performance.
Turning complex models into business tools
Comcast Advertising helps marketers build brand relevancy and achieve sustainable business outcomes through multiscreen TV campaigns. Achieving such a wide reach across a broad national footprint means the organization must analyze massive datasets — from viewership and campaign performance to audience insights, engagement and historical outcomes — to deliver timely insights and continuously optimize their strategy.
Data scientists were able to build sophisticated predictive models, but they struggled to share those insights with business users. Traditional BI tools didn’t provide the flexibility or interactivity needed to let end users explore model outputs in real time. Additionally, building custom applications on public cloud and on-premises systems required extra tooling, hosting overhead and introduced unfamiliar tech stacks. This combination of inefficiencies slowed data science productivity, frustrated teams and limited the end user perspective.
Jivitesh Poojary, Lead ML Engineer at Comcast Advertising, articulated, “We wanted a way to let business users interact with model outputs directly, without requiring them to understand the underlying data science. That meant we needed a platform that let us quickly build applications, customize them and integrate with our data pipelines.”
Comcast Advertising needed a solution that would let them convert predictive models into user-friendly tools, create fast feedback loops and reduce the time from model development to business value. That need ultimately led the data science team to Databricks Apps.
Accelerating app development and collaboration
The data science team at Comcast Advertising was already leveraging the Databricks Data Intelligence Platform for Lakeflow Spark Declarative Pipelines, ML experimentation and governed data access. When Databricks Apps became available, it offered a natural extension of existing workflows, providing a direct way to bring models to life through instinctive, dynamic and interactive applications.
With Databricks Apps, Comcast Advertising found the agility their data teams were seeking. Instead of navigating separate hosting environments or building unfamiliar front-end interfaces, data scientists are now using Python-based frameworks directly within their current development stack. Not only does this lower the learning curve for developers, but it also dramatically reduces time to market. “The ability to build these applications using popular Python frameworks while seamlessly integrating with the broader Databricks Platform creates a powerful environment where data insights can be packaged into easy-to-use interfaces for broader organization use,” explained Poojary.
One of the organization’s first use cases was a forecasting dashboard that allows business users to modify key input levers and instantly see how those changes affect revenue predictions. This hands-on experience gives sales, marketing, strategy and customer experience teams new visibility into model outputs. Furthermore, they can explore multiple what-if scenarios without relying solely on the data science team for every update.
Using lakehouse architecture, Unity Catalog for secure access control, MLflow for model tracking and serving, and SQL Serverless for fast, governed queries, Databricks Apps provides Comcast Advertising with a unified and streamlined path from raw data to real-time decision-making.
Delivering faster insights to directly impact revenue
Using Databricks Apps to package predictive models into interactive tools has shortened the path from data to decision-making for Comcast Advertising. Business users can now engage directly with model outputs — testing inputs, exploring potential downstream impacts and adjusting ad strategies in real time. This shift has enabled more responsive, data-driven decisions across campaign planning, sales and customer experience.
Databricks Apps have also improved operational efficiency for Comcast Advertising data teams. With development centralized in the familiar Databricks Platform, the organization has seen a 10 to 30% reduction in development time, resulting in faster iteration cycles, quicker feedback from end users and increased confidence in model adoption. Poojary remarked, “We’ve saved time, but more importantly, we’ve improved the quality of what we’re building by keeping everything within a unified platform.”
The forecasting dashboard is currently undergoing user acceptance testing, with plans to expand access to campaign managers and scale to support hundreds of users across the organization. Meanwhile, the team is actively exploring GenAI capabilities, like building assistant-style chatbots and preparing to expand their data product offerings.
Looking ahead, Comcast Advertising plans to explore Databricks Lakebase architecture to combine operational, analytical and AI data more easily. This will allow the organization to scale data products and drive future Databricks Apps development as more business teams adopt predictive tools. Poojary concluded, “As we continue to scale our data products, we look forward to simplifying our data architecture with Lakebase.”