Future-proofing ad sales: A modern architecture that ends vendor reliance

A media leader wanted to restore ownership of audience data.
About
Our client, a leading media and entertainment company, operates one of the largest portfolios of film, television, and streaming properties. Its advertising and analytics teams support hundreds of global brands through advanced audience insights, sales strategy, and cross-platform campaign measurement.
Weekly to daily
forecasting, accelerating sales and planning decisions
Months to hours
of data refresh cycles, improving trust and relevance
Multiple times per day
forecasting and reprocessing, increasing agility
Challenge
The client needed to break free from heavy dependence on a single audience-measurement provider—a dependency that created high costs, rigid workflows, and limited negotiating power. Their legacy tooling was brittle, slow to update, and deeply intertwined with proprietary vendor data, making even small changes risky and time-consuming.
Compounding these issues, the business increasingly required rapid data delivery to empower sales, planning, and analytics teams to act immediately at any decision point. But the existing system couldn’t keep pace: slow refresh cycles and delayed updates meant teams were making high-impact decisions using stale or incomplete information.
After years of poor service, inflexible processes, and mounting pressure for faster, more responsive insights, the client sought a modern, provider-agnostic platform that they could fully control and maintain themselves, without sacrificing accuracy, speed, or the ability to react in real time.

Approach
Building on prior success modernizing another part of their data ecosystem, our team collaborated directly with planning, inventory, and data engineering teams to assess workflows, stabilize existing pipelines, and design a modern architecture that could scale, adapt, and be maintained internally.
Solution
Instead of patching an outdated, tightly-coupled codebase, Lumenalta rebuilt their core audience and forecasting tools—from long-term quarterly predictions to log-optimization workflows and upstream data pipelines—using a clean, modular, data-provider-agnostic architecture. Our engineers and ML specialists delivered maintainable, transparent, multi-currency systems powered by modern orchestration (AWS Step Functions, Argo) and standardized ingestion frameworks.

The result: the client now operates a modern, provider-agnostic platform capable of daily forecasting, rapid refresh cycles, and seamless data-provider changes—renewing flexibility, lowering risk, and strengthening their negotiating stance. What was once a weekly process now runs multiple times each day.
Key Highlights
- A fully data-provider-agnostic architecture that reduced vendor lock-in and improved flexibility
- Daily (instead of weekly) predictions supporting faster planning cycles
- Data refreshes accelerated from months to hours, enabling timely, trustworthy insights
- Modernized ingestion pipelines and standardized processes across teams
- Faster, more reliable data reprocessing and analytics, reducing operational risk
- A containerized workflow design that made infrastructure shifts seamless
- A parallelized processing architecture that splits large workloads across multiple machines, dramatically speeding up forecasting, reprocessing, and analytics while keeping systems scalable and responsive
Platforms
- AWS
- Snowflake
- Sagemaker - data science experimentation, model training and prediction, ML, MLOps, and parallel inferences
- Argo - workflows, GitHub Actions
- ArgoCD - CI/CD
- Kubernetes
- React
- Postgres
Capabilities

- Validated and trusted data through transparent, maintainable pipelines
- Data-provider-agnostic tools that reduce vendor lock-in and improve negotiating power
- Seamless operations powered by standardized workflows
- Shared terminology and alignment across teams, speeding collaboration
- Accelerated ingestion and reprocessing, supporting near-real-time decision-making
- Faster predictions and updates, enabling more responsive ad sales planning
- Shortened development cycles that allow product teams to deliver new features every sprint, directly improving user workflows and addressing long-standing feature requests
- Adoption of modern ML practices to isolate the data science workspace from the data engineering space.





