Leveling up By Aliens’ data organization

At ByAliens, data lies at the center of their business model. It is used to manage user acquisition, ads monetization, and product optimizations. It is crucial for the success of a mobile gaming company. To excel in those tasks, they need a robust data organization that serves the internal users’ needs end-to-end.

We collaborated with ByAliens to build the necessary infrastructure for high-quality data for the Marketing and Finance teams. ByAliens' CEO, René Retz, stated “ I knew we couldn't just continue working with the processes we had. We needed to completely change the way we were manipulating data and the team could definitely use some help” . The project encompassed both data infrastructure and data flow, with the objective of removing the current data engineering bottlenecks and delivering a superior set of metrics to the business teams.

We encountered a disorganized data process and data that was arranged in a challenging way. Multiple parallel processes for the same data resulted in inconsistent outcomes and hindered the ability to adapt the data model to incorporate new information the business collected.

We tackled the problems by revamping each step of the data lifecycle. We started by interviewing business users and listing all the metrics they desired and their requirements for freshness, granularity, and slicing and dicing. Then, we improved the connections with their data sources, such as AdNetworks, Attribution Platforms, Google, and Apple, to gain a deeper understanding of the data's meaning and how it could be used to enhance each other. At the same time, we eliminated tool redundancy by migrating event collection from three different third-party software to a single one with standard identification, enabling deeper and more meaningful data connections. 

We then established a layered data model comprising of four layers: raw, cleansed, trusted, and refined. The previous data model had 34 interconnected tables and lacked several industry-standard practices: tables were rewritten in their entirety daily, there was no use of partitions or data clustering, and the table schema did not encourage exploration, only a set of predefined metrics, filters, and groups. To address this, we rebuilt the data model from scratch to organize not only the table structures but also to provide clearer business meaning. This resulted in 17 new tables that processed all the information in an organized manner. Troubleshooting and maintenance became much easier. Our improvements include a data model that encourages data exploration by linking it to business concepts and that can be easily updated to include new data and metrics; a suite of data validation and alerts that ensure data quality; and tables partitioned and clustered according to data consumption patterns, providing lower costs and better performance. Data processing costs to deliver the data dropped from a couple of thousand dollars a month to less than $100.

We completed the project by rolling out the new pipelines to the organization. We quickly received feedback based on actual usage, allowing us to fine-tune the data lake.

The Marketing and Finance teams can now leverage data to its fullest potential to optimize campaign performance and improve the accuracy of accounting processes. This directly relates to an increase in user acquisition, revenue generation, and product development. As René stated:

“We spent two years trying to improve those pipelines with no success. With deX, we managed to do it in 3 months. Now, I feel I have the business under control and the business teams are happy to have support to develop their own projects.” .