SITUATION​

While servicing our client’s high transaction player data pipelines, Everforth Apex analysts identified a value opportunity within the client’s existing infrastructure. Everforth Apex’s team ideated with the publisher to design and develop a function referred to as “Friend Recommender.” This innovative tool automates the creation of new connections on the platform, matching users to foster a community-centric network of engaged players.​

The AI used to automate best-match priority requires data from disparate systems, which caused roadblocks with quality of the friend results, as well as low access speeds and large system compute burden. It became quickly evident that to make the systems both usable and sustainable, a custom data solution would be needed for the viability of the product.​

Facilitating the integration and deriving insights required us to extract, transform, load (ETL) and unify the data in a cached and optimized manner to allow non-disruptive centralized querying while limiting the impact of data comparisons on the overall systems. Neo4j was chosen as an intermediary data system to produce connections between data existing in various systems and to function as an insight store as well as query cache to the interconnected data stored on largely GCP-deployed Databricks stores. ​

Everforth Apex’s data architects designed, documented, and engineered a solution centered around the graph database with an acute focus on analytics and reduction of overall system impact. ​

RESULTS ​

• Net average 15 million daily friend recommendations​

• Daily Player churn reduced by 22%​

• 60% friend acceptance rate ​

• ~17,000 new daily friend requests with an average 10,000 new gaming connections​

• ~$1.04M daily revenue retention due to Player churn reduction.​

• Negligible net cost increase on solution team resources and net computer cost​