Rearchitecturing and Ongoing Enhancement of Demand-side Platform

With the old architecture, it took launching a separate application instance to register and manage each client, which led to enormous resource consumption and poor scalability. The process was cumbersome as the administrator had to manually load a static config to the server and then edit it, if required. To solve this issue, the project team rebuilt the platform with the following outcomes:

One application, one database. All client entities are now stored in a single database and managed via a single application, which saves server resources and optimizes their usage.

Admin panel. New clients can be created and managed from a single admin panel in a few clicks. The broad scope of the panel functions allows editing clients’ details (name, logo, geotargeting, and more), adding user accounts and defining their rights, assigning inventory types, and managing third-party data providers for highly specific, precise audience targeting.

Depending on their rights, end users get access to an extended, optimized functionality that allows to:

• create and manage laser-targeted marketing campaigns that incorporate sophisticated behavior-driven ad delivery
• deeply customize ad campaigns, including media plans, start and end dates, inventory sources, budgets and daily spend, bids, third-party data providers, and more
• upload and manage multimedia creatives such as video and images
• generate highly customizable reports using a variety of metrics and filters (advertisers, data providers, inventory research, publishers, conversion pixels, RTB, and more)

To deliver continuous and smooth reporting, the system employs multithreaded data processing when exchanging large data volumes with a RTB platform. Multithreaded processing also ensures no time-consuming data migrations or backups can interfere with the database performance and user workflow.

In addition, Oxagile’s team took an active part in creating new features and extending existing ones. Among others, this included:

• improved handling of creatives
• dynamic RTB seat management
• advanced reporting with new metrics and totals calculations, as well as new targeting options
• tracking of changes made to client entities
• batch editing of campaigns and creatives

To minimize regressions when dealing with business logic of this complexity, the project team launched a dedicated automation server that performs over 3,000 unit tests several times a day, with 80% of test scope coverage. The codebase also greatly benefited from regular internal cross-reviews prior to committing.