THE CHALLENGE
Scaling analytics beyond a high-maintenance legacy data stack.
Every ad Flashtalking served generated log data, and the company had ads running across the internet around the clock.
That data had to come together from logs on every server before the team could build the reports their advertising agency clients depended on.
Before working with DAS42, Flashtalking’s engineering team was running that process on a mix of homegrown tools and open-source MySQL scripts – tools built for transactional workloads, not analytics. The gap showed up in two places: the team had to stay on call 24/7 to keep customer service where agencies expected it, and the legacy stack was getting harder to maintain and scale as volume grew.
We are processing data pretty much flawlessly now! … This is transformative for us.
– CTO, Flashtalking
The Solution
Migrating without disrupting the business.
DAS42 started with the stack Flashtalking already had, prioritizing one thing above everything else: protect the current data and keep new data flowing in without a gap.
After discovery, DAS42 built a parallel platform alongside the legacy system. Over two months, DAS42 tested the new platform against the old one to confirm it produced the same results before cutting over. Running both systems side by side let the team QA the new platform against real production data while keeping the business running at full speed.
Airflow and Snowflake as the new foundation
DAS42 replaced the homegrown tools and transactional MySQL scripts with Apache Airflow and Snowflake. Airflow pulls data from every server and centralizes it; Snowflake holds legacy data and everything generated since in one place built to scale with the business.
The results
Greater Visibility, Scalability, and Operational Efficiency.
With the new platform in place, Flashtalking’s team has full visibility into every stage of its pipeline, fewer vendors to manage, and a system that’s easier to run day to day.
That matters most in Q4, when retail clients ramp up ad spend and servers run closest to capacity — the new platform is built to handle that peak load instead of straining against it. It also gave Flashtalking a stronger foundation to integrate into the Mediaocean stack following the acquisition.
We work as an extension of our clients’ teams, solving the data problems that stand between them and a business decision.
Real clients, real results.
