Neue Zürcher Zeitung Builds a Data Ecosystem That Supports High-Quality Journalism

For more than 200 years, Neue Zürcher Zeitung (NZZ) has delivered German-language news from its headquarters in Zurich and its 40 bureaus around the globe. While print remains an integral part of its offering, NZZ has expanded significantly into digital formats, including web, video and podcasts. One thing has remained constant: a commitment to high-quality reporting supported primarily through a subscription model.

Delivering news and reporting that keeps subscribers coming back is no easy feat, and it only gets more challenging when you can’t get the latest insights into subscriber preferences and behaviors.

NZZ quickly found its on-premises SQL server limited its ability to combine commercial and user-behavior data — preventing it from truly understanding its readership. To get the fresh insights it needed, NZZ deployed Snowflake’s AI Data Cloud.

NZZ’s legacy SQL server warehouses handled only commercial data, such as subscription numbers and customer relationship management system entries. As its business and operating model has evolved though, NZZ’s data team realized it needed more insight. “To keep customers loyal, we need to really understand their tastes and behavior,” explains João Soares, Head of Data at NZZ. “We needed a lake to store and analyze larger volumes of user-behavior data collected from our app and website. However, our SQL servers were struggling with the higher transaction levels.”

As it went from processing hundreds of thousands of records to millions, NZZ began looking for a platform that could support this increased load and truly combine this user information with transactional data — while working well with semi-structured data and more complex schema.

NZZ primarily uses Google Cloud architecture, so when it began running proof-of-concept projects, it started with BigQuery. However, after testing BigQuery alongside Snowflake’s AI Data Cloud, NZZ found Snowflake came out ahead. “What really set the Data Cloud apart was its ease of use and high maturity; everything just worked exactly as described,” says Soares.

As NZZ began implementing Snowflake, it made the decision to start with a greenfield approach. By getting rid of all legacy code and technical debt, the company has been able to create a simpler data ecosystem, combining the AI Data Cloud with dbt for transformations, Meltano for ingestion, Airflow to assist with orchestration and Tableau for reports and dashboards.

In just two months, Soares and his team were able to deploy the initial databases and warehouse, integrate Snowflake with NZZ’s single sign-on authentication, implement user access and start ingesting data. “Data is always a journey. Our destination is to have Snowflake as the center of everything,” Soares says. “Around 90% of our data is already going through Snowflake. Everything from behavior insights to article content and metadata: It’s all there.”

In the earliest steps of its Snowflake journey, Soares and his team embraced the move to a simpler platform in the cloud. “Our developers really appreciated saying goodbye to some of those old SQL server headaches,” recalls Soares. “No more discs filling up or memory running out. And no server hiccups every other month.”

The flexibility of the AI Data Cloud also accelerates development times. NZZ teams are able to easily provision and resize warehouses as needed to test new use cases and data products.

In addition, Snowflake’s zero-copy cloning features have simplified NZZ’s deployment pipeline. In the past, deployment was a largely manual process that required executing SQL and Python code against production databases — an error-prone method that could lead to unplanned downtime. With Snowflake, NZZ built a blue/green deployment process where new code gets executed against a cloned production environment, which is switched over when the execution and tests are successful. This creates a faster, more flexible, more cost-effective deployment process that avoids disruption to business processes.

The AI Data Cloud’s extensive integrations also helped NZZ be more flexible with how it collects and analyzes user data. When NZZ began using Piano Analytics to track user behavior data, it was able to share Piano databases directly in Snowflake to get data in near real time.

With Snowflake as the core of its data ecosystem, NZZ can now refresh more data, more quickly and at a much lower cost. Today it stores 13TB of data in Snowflake, as compared to the single terabyte it could fit on its SQL server, and that data is refreshed hourly instead of once per day. A more efficient data engine and reduced admin requirements also cut costs by 40%.

Beyond NZZ’s four walls, a key benefit is that a beacon of journalistic quality and integrity can continue to serve its readership with the very best news, opinion and analysis. “Journalism is only important if people read it,” says Soares. “We can now use data to really understand what our readers think is important. That helps us provide valuable feedback to our editorial team so they can keep delivering high-impact journalism.”