Detection of Impression Losses in Sklik Ad Serving Zones for Seznam.cz
Sklik’s advertising ecosystem is so vast and dynamic that even a brief outage in a single serving zone can result in unnecessary, often hidden revenue loss. Seznam.cz therefore needed a solution capable of automatically detecting abnormal behavior on a large scale—particularly sharp drops or zero impressions—and alerting them in a timely manner.
At Data Mind, we designed and delivered a custom tool that evaluates thousands of zones across devices, browsers, and ad types using historical data, determines the severity of anomalies, and provides clear outputs for rapid operational remediation.
Data Science
We discover hidden patterns and connections in your data for better customer understanding, process optimization, and future development prediction.
Process optimizations
Sjednocení a integrace dat z různých zdrojů, automatizace, zrychlení zpracování dat...
Data Engineering
We build reliable and scalable infrastructure for data management that ensures data integrity, availability, and efficient processing.
Web analytics
We measure and analyze the behavior of visitors to your websites and applications for better performance, conversion, and user experience.
Document Automation
data extraction, summarization, classification, manual-free workflow
Key findings
Automatic detection of impression failures
across nearly 10,000 delivery zones daily
Significantly faster identification and resolution of incidents
which reduced ad downtime
Better protection of Sklik's advertising revenue
thanks to timely alerts about anomalies
Would you like similar oversight of your data or systems?
Schedule a consultation
Goal
- Identify impression losses
- Report anomalies
- Minimize hidden losses
- Speed up the resolution of issues
- Support stable income
We identified the ad impression outages in time to ensure that Sklik revenue would never drop for an extended period again.
Advertising is a key source of revenue for Seznam.cz, and it is distributed across many sales channels. The system operates with a vast number of ad zones, the volume of which can no longer be reliably monitored manually. Although Seznam.cz was already using basic monitoring tools, it lacked a targeted analytics layer capable of reliably detecting ad display failures in specific zones and alerting the team in a timely manner.
Key Challenges
- 10,000 zones per day
- Two years of history in daily resolution
- Different standards of conduct in each zone
- Selection of appropriate algorithms
The key challenge was scaling the system to handle thousands of time series while accounting for differences between zones in terms of size, equipment, and other parameters, to ensure the system did not generate false positives.
We’ll help you identify data issues before they impact your businessWe’ll examine your data, systems, or marketing platforms and propose solutions that ensure timely problem detection and consistent performance.
Data Mind Solutions
- Deployment in the Keboola cloud infrastructure
- Definitions of anomalies and their severity
- Filtering out inactive zones
- Merging small zones
- Voting among multiple statistical methods regarding the severity of an anomaly
- Different alert thresholds for small vs. large zones
Technologies Used
- Keboola Connection
- Python detection pipeline
- MissingDays check
- Forest Insulation Detector
- Prophet Trend Monitoring
First, we profiled the data, calibrated the definitions of outages and their severity, and reduced the number of zones through smart aggregation. We then selected a combination of algorithms based on the behavior of each zone type.
Result
The solution provided global reporting on delivery zones and an overview of Sklik’s delivery performance across the system. It also includes a set of detected anomalies with automatic alerting, which enables rapid identification of issues and their escalation to operations. We delivered both summary reports for analytical work and machine-readable anomaly outputs, which serve not only to escalate outages but also as a practical tool for identifying unmaintained or unused zones.
Benefits
- Instant detection of outages
- Shorter incident duration
- Fewer undisclosed losses
- Greater operational reliability
- Growth in advertising revenue
Thanks to timely alerts about ad impression outages, the time between the onset of a problem and its resolution was reduced, which led to shorter incident durations and helped protect advertising revenue.
Are you facing a similar issue with data or system monitoring?
Discuss the optionsWould you like similar oversight of your data or systems?
Want to go deeper? Dive into our knowledgebase
Customer 360
Creating a unified Customer 360 profile is not just a marketing goal, but a complex architectural challenge. This article analyzes the necessary transition to Data Lakehouse architecture in the Microsoft Fabric and Azure ecosystem. Learn how to build a modern data platform that serves as a robust foundation for advanced cloud analytics and hyper-personalized communications using AI and LLM.
RAG over company data
Want to deploy generative AI on top of your company's own know-how without the risk of hallucinations? This technical deep dive into RAG architecture shows that AI success stands and falls with quality data engineering and data architecture. Explore how to build a modern data platform for AI based on Microsoft Fabric and Azure Data Lakehouse and how to overcome the challenges of data vectorization, security, and transition to a production enterprise environment.
Propensity to Buy
How can raw data be transformed into accurate predictions of purchasing behavior? The Propensity to Buy model represents the gold standard of data monetization, but its success stands or falls on the chosen architecture. In this article, we will explore the technical background of implementation in Microsoft Fabric and Azure environments. Learn how to leverage Data Lakehouse principles, advanced machine learning, and generative AI to identify customers with the highest potential—and how to orchestrate it all in a modern, scalable infrastructure.