Audacy is a media company that owns 235 radio stations and a portfolio of podcast channels across 48 media markets. It distributes content via branded streaming apps over the web, mobile, and smart home devices, as well as traditional over-the-air radio. Like most media companies, the revenue model consists of the delivery of advertising to segmented audiences across its media platforms.
- Current data stack: Snowflake, AWS, SSIS, GCP, Airflow, and Python
- Main data sources: Oracle, Nielsen, Wideorbit, Adswhizz, Tapclicks, and various DMPs (data management platforms)
Team: 3 – 5 analytics engineers
Reduce the number of tools to a single platform:
Accelerate delivery from 45 days to 3 days
Deliver small data requests in under an hour
Raise productivity of our small team to move faster and get more done
Automate change propagation:
Eliminate pipeline outages that stop business from getting data it needs
Eliminate lost data that used to span months before it was detected
Instant error detection drives issue resolution to under an hour
Partnership for innovation:
Migrate 75 data projects from legacy code in a few weeks
Onboard analytics team by working side-by-side with Ascend engineers
The Brittle Data Pipelines Problem
Terry Dougal joined Audacy as Data Protection Engineer as it transitioned from local databases to using Snowflake. “Snowflake greatly reduced our DBA overhead and simplified our infrastructure,” said Terry in a recent interview with the Ascend team. “But we also had over 75 projects orchestrated with Airflow and a combination of tools, cloud services, and bespoke code. The system suffered frequent outages and regularly lost months of data. We couldn’t get data to the business fast enough, and the data we did ship wasn’t trustworthy. Our velocity of delivering new data products continued to diminish. This limited the value of analytics for the business, and led to lengthy projects to fill the gaps.”
Intelligent Data Pipelines at Audacy
Instead of hiring more data engineers to patch over all the tooling with more bespoke code, Terry shifted gears and looked for a new solution. “I knew there had to be something out there that automated the manual things we were writing and supporting. We evaluated Matillion, Pentaho, ETLeap, and dbt, and ultimately selected Ascend as our Data Pipeline Automation platform,” said Terry.
“Ascend is simply way more than just an ELT tool. It won with ease of use by our analysts, the comprehensive scope of its automation, and the speed of end-to-end pipeline development. The Ascend team also really partnered with our data team, and they are teaching us the platform as we go. The initial wave of work involves the migration of 75 data projects, which is proceeding like clockwork!”
Ascend has helped put the strategic business goals of Audacy within reach while avoiding a massive push to hire data talent. This is because Ascend automation is relieving the team from worrying about system reliability and orchestration complexity, which in most organizations take up to 90% of the team.
Creating Business Impact
Audacy is shifting its business growth strategy from acquisition of station and podcast networks to creating value within those networks through advertising and audience growth. At the heart of this strategy are new data products, including audience analytics, content analytics, and advertiser analytics. These products have sophisticated data pipeline requirements, involving constant tuning and change. They also put unprecedented demands on the data team, including velocity to build and launch new data products, reliability of the pipelines, and quality of the data products.
Terry now leads the data team to prepare for this future. For example, Audacy uses external managed data providers to assist in segmenting their audience. As Terry describes, “Audacy itself has unique usage data based on the listening habits of our users. This includes which content they are listening to when, what their dwell time is on specific content, which ads they engage with, and more. With Ascend in play, our small team can now develop sophisticated new segmentation data products that will not only match listeners, content recommendations, and advertisers directly on Audacy’s own digital platforms, but we can also market across our portfolio of radio stations and beyond.”
Increasing the number of podcasts available on Audacy platforms is a strategic goal. To compete for and promote podcast talent, Audacy data products will be able to identify emerging content leaders across the expanding universe of niche audiences, and support their monetization goals with hyper-targeted advertising.
Audacy approaches data pipelines with automation as the organizing principle. This is enabling a holistic view of how infrastructure, automation, data sources, external providers, and internal business partners fit together to create new analytics-based capabilities. By consolidating the functions needed to build and operate data pipelines at scale, Ascend automation unlocks this new level of efficient and business-oriented DataOps.
As Terry tells it, “we are implementing a workflow by which any business unit can submit a Zendesk ticket for a new data product. Some requests may start with a simple data movement request to get data from one application to another on a regular basis. Our team can prioritize and fulfill such small requests within a few days, after which it runs reliably. One such legacy data flow had to be triggered manually and took two days to run – on Ascend, it now completes on schedule and in under an hour!”
The ancillary benefits of a unified pipeline automation platform are dramatic. Since the new data flow is now fully automated, orchestrated, and instrumented with Ascend, the data team is instantly notified of any data irregularities. With just a few clicks, it can also include the business on specific system notifications, and reach out proactively to resolve any impact.
With its comprehensive approach to automation, Ascend raises the visibility of every data source, how it is used and in which data product, and the reuse of data products across the business. This complete end-to-end lineage is visible in real-time, making it easy to ideate with the business and pilot new data products quickly. Velocity improves not just in data engineering, but in the entire data product life cycle, from idea through to robust operation in production.