reduction in time to insight
The Canadian Broadcasting Corporation’s (CBC/Radio-Canada) mission is to inform, enlighten and entertain their vast and diverse pool of listeners. One of its core strategies to achieve this goal is to personalize their digital broadcast radio services. To accomplish this, they sought to extract insights from their data to better understand signals such as subscriber trends around churn, content consumption and relationships between different types of content to drive stickiness.
“We’re trying to pivot to a more data-informed organization. That means giving more visibility over our digital audiences and their behaviors in a self-serve manner,” explained Stephane Caron, senior director of business intelligence at CBC/Radio-Canada.
Like many businesses that hope to drive more engagement on the promise of personalization, CBC/Radio-Canada ran into issues with their legacy Hadoop infrastructure, which was highly complex and costly to scale. Not only did teams struggle to access data, they also had been finding it difficult to explore the data, which included where and how people access content (geolocation, types of devices, etc.), audience profiles and interactions with various digital products (programs viewed, content clicked on or shared, etc.). Further slowing the teams’ ability to extract actionable insights was the sheer volume of data being generated, pushing them up against their computational power limits. Things got even more complicated on Hadoop — particularly with stability and cluster maintenance — as more data started to pour in.
“The rigid nature of our Hadoop system meant we had to share the same resources and manage them carefully to control costs, which limited us in our ability to run parallel analysis on a large volume of data,” added Caron. “We’re a small team that cannot afford to have a full-time team of DevOps to maintain and monitor our clusters. We knew we needed to migrate to a cloud-native platform in order to remove that pain point.”
Once CBC/Radio-Canada migrated from Hadoop to the Databricks Lakehouse Platform on their Azure cloud, their data went from barely to easily accessible, which led to a ripple effect of benefits.
For starters, with the adoption of a lakehouse architecture, all of their data was immediately at their fingertips. Now data engineers are able to quickly build data pipelines that feed into visual dashboards via Power BI faster. And with Databricks SQL, they are able to quickly run SQL queries from Power BI to update their dashboards, reducing the amount of time it takes to derive insights into their digital audiences and their behaviors — from weeks to minutes. And with optimizations to provide the best performance on all query types, they are able to experience superior price/performance compared to their legacy data warehouse.
“Databricks Lakehouse combines the advantages of a data lake with those of a data warehouse, where you can use asset transactions like deletes and merges and simply interact with tables as if you were in a real data warehouse,” added Caron. “This has unlocked a tremendous amount of value within our data that we weren’t previously able to leverage.”
Delta Lake provides CBC/Radio-Canada with a common data layer to build reliable and scalable data pipelines with ease. What this does is bridge the gap between data engineering and data analysts, allowing teams to access and leverage all their data in the ways they are familiar with. And with ACID compliance, data reliability and consistency are no longer questioned.
Caron explained. “Now, the analysts actually trust the data. Before, we had to check in daily just to make sure that what was in was actually reliable. Now we’re confident that it is, and that’s made a big difference in our ability to collaborate effectively. Data confidence gives us the ability to build ETLs much faster, which means we’ve automated a lot of our responsibilities, freeing up time to focus on more important things.”
With accessible data, near real-time insights and infrastructure simplicity, operations have overall been streamlined. The CBC/Radio-Canada team can now easily provide their data to various data consumers to help create differentiated services that engage and delight listeners.
Today, with the help of Databricks, CBC/Radio-Canada has cut their time to insights into audience behavior in half — if not more. On their legacy Hadoop technology, they were very limited by the volume of data it was able to work with, but now the scale at which their platform can operate is no longer a concern. With these insights, CBC/Radio-Canada is now able to provide more visibility into their digital audiences and develop strategies and services that will boost engagement and retention.
As far as their future with the platform is concerned, the data teams are ready to explore what else they can do in terms of personalization, now that they have the data access and scale that they need. “Personalization is the key pillar in our five-year strategy,” said Caron. “Databricks will play a key role there, as it’s the underlying infrastructure and the underlying platform we’re using to derive insights out of our data.”