Advancement of ad technology by utilizing data, AI and machine learning
Monthly cost reduction compared to the existing Hadoop system
Reduction in processing and collection time for ~100TB of data per month
Reduction in data analysis/modeling implementation time
In the fast-changing digital advertising market, advertising media and techniques are becoming more diversified day by day. In particular, ad technology that maximizes advertising efficiency using artificial intelligence and machine learning is attracting great attention, and MezzoMedia, an integrated digital marketing company, is leading the digital advertising market by strengthening its data and technological competitiveness. MezzoMedia has shown consistent growth since it took its first step as the first digital media rep company in Korea. Currently, as a company with the largest digital partnership in Korea with its own advertising platform and solutions, MezzoMedia is providing a variety of specialized digital marketing solutions. Since its foundation, it has executed campaigns for 1,600 advertisers, and its billing volume has steadily increased, reaching 620 billion won last year. However, several limitations imposed by the existing on-premises Hadoop system hindered its growth. To overcome this, MezzoMedia introduced the Databricks platform, which can support more flexible and diverse workflows. With the introduction of Databricks, MezzoMedia has been able to explore, analyze and monitor large-scale data with higher efficiency, and improve collaboration between teams to significantly increase overall data management productivity through changes in data linkage, individual data management and authority control.
MezzoMedia’s digital transformation: from on-premises to cloud
The priority for MezzoMedia is creating new value and achieving innovation by breaking free from the existing framework utilizing data-based technology. MezzoMedia manages an integrated data pipeline using data, artificial intelligence and machine learning, and aims to improve the data literacy capability of all employees. Furthermore, MezzoMedia seeks to develop a model for identifying and predicting patterns and trends in data in order to minimize the time and cost of data processing and analysis, and to adapt to the changing market.
However, the existing on-premises Hadoop system had several resource limitations and restrictions in realizing this business model. It was difficult to implement various workloads due to the complex infrastructure and high license costs; data processing speed was slow, as flexible scaling was infeasible; and usability was poor, as simultaneous query execution was infeasible. In addition, due to restrictions on data access management, data silos were formed between teams, resulting in the inevitable reduction of data quality and consistency.
Building the data pipeline was also a challenge. Hadoop’s complex data pipelines were built on different solutions or servers, hindering critical processes such as data flow management, scheduling and error handling. There were also problems with data search and metadata management. The process of finding the necessary data was not easy, as data was distributed and stored by solution, and it was difficult to efficiently operate the infrastructure necessary for large-scale data processing in an on-premises environment. To overcome these difficulties, MezzoMedia felt the urgency of introducing a cloud platform that could collect and analyze data more efficiently and accommodate various workflows.
Efficient data management through Databricks’ integrated solution
With the introduction of Databricks, MezzoMedia is now able to manage large amounts of data with efficiency in a single, integrated solution. Instead of going through multiple services and environments, data collection and analysis are possible on a single platform, simplifying the existing complex processes and workflows. Data engineers can respond more actively to a variety of use cases by performing external/internal data integration and setting individual data management and account privilege controls. Meanwhile, data analysts can improve productivity by leveraging the integrated system that scheduled and automated routine data processing and analysis tasks to reduce repetition. It also enables the large-scale data processing required to develop machine learning and deep learning models for ad anomaly detection and performance prediction.
Most importantly, with the introduction of Databricks, 70%–80% of the monthly cost has been reduced, in comparison to costs of the existing Hadoop system. Furthermore, as no additional server installation or management work was required, operational resources were reduced, and resulted in a significant reduction in business operation cost. In addition, stable service operation improved system stability, providing customers and users with high data reliability. With these advantages, MezzoMedia now has an environment that can further expand cost savings in the long term and respond flexibly to ever-increasing data. Furthermore, Databricks has played the intermediate role in reporting advertising performance data to clients in an easy and quick way, which has had a positive impact on indirect sales.
Improved collaboration through leveraging the Databricks platform
With the introduction of Databricks, MezzoMedia has simplified and democratized data access issues that were limited in the existing system. As a result, Databricks’ integrated platform facilitated collaboration not only within the data team, but across all internal organizations. It is no longer necessary to provide data in separate file forms, making it easier and faster to share data. Users can easily view and analyze tables created by engineers, enabling smooth communication and maintaining data consistency. In addition, by using the table or column comment function, non-developers can also easily understand the data information. This has created an environment where administrators can spend less time on support for users and more time on operational and management tasks. Non-experts could also obtain essential data-driven insights through Databricks’ simple interface and make data-driven decisions accordingly, which would increase overall business efficiency.
Strengthening ad technology competence by introducing Databricks
Through Databricks, MezzoMedia has significantly improved productivity in data management and collaboration within the organization. Data collection, analysis, modeling and machine learning can all be performed on the integrated platform of Databricks to improve work efficiency, while cost efficiency also increases as operation and management become easier with a simple and flexible architecture. As a result, MezzoMedia has made its leap as an artificial intelligence/machine learning-based ad technology company, and the use of various data also increased to drive growth in the business field. Engineer SangHyun Lee of MezzoMedia’s Data Tech Team, said, “Databricks has strengthened MezzoMedia’s ad technology capabilities to maintain a competitive edge in the digital advertising market, and now, MezzoMedia’s foundation has been laid to grow into Korea’s No. 1 integrated digital marketing company.”
