In the modern enterprise tech stack, data is the active engine driving real-time decisions, generative AI, and hyper-personalized customer experiences. But, while application reliability standards have matured over decades, data reliability often remains a wild frontier.
At the previous edition of the Data 2030 Summit, Daniela Kato, Data Management and Governance Lead at Zalando, delivered a compelling look on how Europe’s leading online fashion platform is embedding Data Service Level Objectives (SLOs) directly into their operational process.
Data Incidents Carry a Heavy Price Tag
Zalando operates at a massive scale: over €15 billion in revenue, 50 million active customers in Europe, and a federated tech workforce where 10% of all employees (2,000+ engineers, analysts, and data scientists) directly build and consume data products.
However, Kato shared a look into the true cost of data unreliability:
- The Detection Gap: While application outages are caught in minutes, data incidents took an average of 12 to 13 days to detect.
- The Resolution Delay: Restoring application services takes hours; repairing complex, upstream data pipelines often takes days of deep root-cause analysis.
- The Financial Impact: Undetected data failures resulted in double-digit millions of euros in lost revenue in a single year, impacting more than 50 teams.
For an e-commerce powerhouse driving cutting-edge features like the Zalando Fashion Assistant (powered by GenAI) and hyper-personalized search ranking, bad or delayed data isn’t just an IT nuisance but also a direct business risk.
Applying DevOps Practices to Data (DataOps)
To mitigate operational risk, Zalando is bringing core DevOps principles like automated testing, continuous monitoring, and infrastructure-as-code, into the data realm. However, this talk emphasized a critical baseline lesson: there cannot be implementation of Data SLOs over chaos. 
Before introducing SLOs, organizations can build a solid governance foundation:
- Clear Data Inventories: Knowing exactly what data assets exist.
- Explicit Ownership: Eliminating ambiguity around who builds and maintains data sets.
- Consumer Awareness: Mapping downstream data lineages to understand where unreliability will hit hardest.
Once this foundation is active, Data SLOs serve as the contract between producers and consumers. Zalando structures these around three core layers:
- Data Set SLOs: Focusing on dimensions like freshness, completeness, validity, and accuracy (e.g., ensuring a machine learning search-ranking data set is 99% available by 5:00 AM with strict error budgets).
- Data Pipeline SLOs: Measuring job completion times across multi-product pipelines.
- Event-Based SLOs: Guaranteeing low latency and high availability for real-time stream processing.
Governance with Teeth: The Tiering Framework and Human Process
Not all data is created equal. To avoid operational burnout, Zalando mandates a tiered policy:
- Tier 1 (Mission-Critical): Mandatory SLOs with formal error budgets and governance enforcement.
- Tier 2 (Important): Opt-out model (teams must formally justify not having SLOs).
- Tier 3 (Exploratory/Local): Opt-in model.
Crucially, Zalando recognized that software monitoring tools alone aren’t enough and the human process is key. Zalando elevated data to a first-class citizen by integrating Data SLO performance directly into their existing weekly operational review meetings, analyzing application and data reliability under the exact same lens.
Unlock the Full Strategy at Data 2030 Summit
Daniela Kato’s presentation provides a rare, transparent look inside one of Europe’s largest tech success stories, detailing their exact 3-stage implementation methodology, mathematical error budget calculations, pilot outcomes, and lessons learned on tool limitations.
The full presentation is currently available exclusively on-demand for registered members and attendees.
Ready to transform your enterprise data practices? Don’t miss the opportunity to join world-class data leaders, DataOps pioneers, and AI strategists in person. Reserve your seat for the upcoming edition of the Data 2030 Summit in Stockholm to gain actionable frameworks, network with industry peers, and drive operational excellence across your organization.
