Masthead Data x VEED.IO
Automating BigQuery Cost Optimization Across Petabyte-Scale Workloads
Achieved within 4 days
~30%
BigQuery compute cost reduction
Data stack
Google BigQuery, dbt, Fivetran, HEX
Site
Industry
AI Tools
Company size
50 - 200 employees
Founded
2018
Masthead gave us instant clarity on where our BigQuery spend was coming from—down to the pipeline and tool level—without weeks of dashboard work. We found dead-end workloads quickly, fixed what mattered, and moved the right jobs to reservations. The best part is we reduced compute costs while improving throughput, which freed budget to invest more in AI.
Bogdan Banu
Founding Data Platform Lead
VEED.IO is an AI Video Creation Platform built to make professional video creation accessible to everyone—used by millions of creators and businesses worldwide. VEED has rapidly expanded its AI capabilities, including advanced AI video editing features, to help users create and edit content faster and more easily.
At VEED’s scale, the platform handles more than three petabytes of video data each month and continues to grow—supported by a lean team that prioritizes speed and minimizing operational overhead through managed Google Cloud services.
Google Cloud is VEED’s strategic partner, providing the foundation for both their core data platform and their AI innovation—from the managed infrastructure they rely on to the AI stack powering new product capabilities.
The Challenge
VEED’s data platform operates at massive scale, processing multiple petabytes of data per month across Google Cloud, spanning multiple projects. Despite that scale, the entire data estate is supported by a lean team of three data engineers responsible for the company’s broader data needs.
With a strong lean culture, the team wanted to direct more of their budget and effort toward AI initiatives and agentic capabilities, not day-to-day data platform cost work. They were aligned on a clear priority: investing in AI and agentic workflows is the future—both for the team’s skill set and for building a more efficient platform over time.
BigQuery was the largest cost line in the data stack, and the team knew there was room to improve efficiency. The challenge wasn’t capability—it was focus. VEED could build internal dashboards, map spend across projects, and run detailed analyses to find optimization opportunities. But doing that would pull engineering time away from the work that mattered most: shipping new AI-driven capabilities and experimentation.
In short, VEED needed a way to identify and execute BigQuery optimizations quickly—without turning their lean data team into a FinOps reporting function. Their goal was simple: unlock cloud budget for innovation and reinvest it into Google Cloud’s AI and agentic stack.
The Solution
VEED.IO already used Masthead for data observability—monitoring data availability, quality, and pipeline performance across multiple Google Cloud projects. This gave the lean data team real-time alerting for pipeline failures and data issues, which is essential at VEED’s scale.
To align with VEED’s broader goal—freeing budget and engineering time for AI and agentic innovation—the team expanded their use of Masthead into Data FinOps observability. Masthead surfaced cost performance and cost attribution across BigQuery in a way that was fast, actionable, and did not require building internal FinOps dashboards.
A recurring pain point for the team was the true cost of ownership of tools in the stack. For example, Fivetran—already known as a premium-priced solution—was also responsible for roughly 30% of VEED’s monthly BigQuery costs.
Pic. The cost of BigQuery compute generated by every solution i team stack. No integration with solutions, no label mapping needed.
Masthead surfaced this attribution within two hours of deployment, without requiring VEED to integrate Masthead into third-party tools or manually stitch cost data across projects.
On the Pic above, Masthead shows, in lineage, BigQuery entities and processes that haven’t been used for the past X days but still consume compute. Typically, these account for 10–15% of total compute in a customer environment. Within hours of deployment, users receive a report with clear cost attribution and a list of unused processes.
Team get a report of this process with clear cost on a pipeline level, which process should be stoped.
After removing these dead ends and validating better refresh frequencies for selected pipelines and dashboards, VEED implemented Masthead packages for dbt and Airflow. This enabled automated optimization of BigQuery execution by assigning reservations to workloads that are more cost-effective on reserved capacity than on-demand.
After removing these dead ends and validating better refresh frequencies for selected pipelines and dashboards, VEED implemented Masthead packages for dbt and Airflow. This enabled automated optimization of BigQuery execution by assigning reservations to workloads that are more cost-effective on reserved capacity than on-demand.
Masthead UI screenshot
On the graff about the customer can understand how much they are paying for the TiB of data processed before and after automatically asigninng reservations to recommended dbt and Airflow models.
The initial result was immediate: VEED started paying ~30% less per TiB processed on average, by applying Masthead’s arbitrage framework and automatically managing reservation assignments.
Masthead identified the best candidate models for reservations using retrospective workload data over a monthly period, simulated workload performance, and proposed a reservation plan with clear max slot allocation.
Masthead UI screenshot
Masthead brings clarity to teams when planning a move to reservations. We don’t just suggest which workloads are the best fit for reservations—we also consider when they run, how models are grouped, queue time, and execution time. This gives teams confidence they will achieve the expected balance of performance and cost after the change, without weeks of manual experimentation.
The Results
By combining data observability with FinOps visibility, VEED redirected effort and budget away from manual cost analysis and back into product innovation—without compromising performance.
~30% BigQuery compute cost reduction, achieved largely automatically within four working days with the effort of one data engineer
More budget freed for experimentation with Google Cloud AI and agentic capabilities, aligning spend with VEED’s innovation priorities
Improved pipeline performance—not degraded. By shifting the right workloads to reservations and optimizing execution, the team effectively added workload capacity while paying less, improving overall throughput and stability.
End-to-end cost visibility across multiple projects down to pipeline/model level—granularity that is not available in the Cloud Console
3 happy data engineers. Major time savings for a lean team: instead of spending months building internal dashboards and doing manual cost attribution to find the most expensive jobs, they saw it all in Masthead and acted immediately.
With Masthead, VEED can continuously monitor data reliability, quality, consumption patterns, and cost in one place. The result is not only stronger data observability, but also Data FinOps observability at any level—helping VEED optimize and control BigQuery spend while keeping focus on shipping and innovation with AI capabilities.
