Our client, the digital division of a Fortune 500 media company, has ridden the wave of OTT (over-the-top) video ads across high-value audiences like YouTube, Hulu, social media and more. The business has grown on the strength of its agency and global brand customers, providing precise geo-targeted audience access to integrated feeds from over 150 ad exchanges and streams, processing tens of billions of ad calls a day.
The Challenge
As the 2020 pandemic lockdown caused digital media consumption to skyrocket, the company set out to open a new market for its SaaS product: precision-targeted OTT video ads by SMBs targeting their local markets. It meant opening a completely new segment of the market. The new product offered the benefits of real-time analytics and bidding management engine — but without the large-scale advertising fees traditionally associated with global-brand-to-agency supply chains.
Building out a tiered self-service offering by extending their multi-tenancy architecture meant they could sell to local businesses at a much more attractive price point. By leveraging the business logic of their existing campaign workflow and management platform, they could meet Time To Market objectives without the need to invest in a massive refactoring of their operating platform. This meant adapting functionality for campaign creation, secure data storage for creatives and marketing data, user data isolation, campaign progress real-time reporting, and budget management.
The Solution
To achieve maximum leverage from their existing enterprise-scale bridge model SaaS platform already running on AWS, the company enlisted CloudGeometry. The goal was to architect and implement microservices in a more flexible shared multi-tenancy strategy known as the “pooled model”. It also delivered three key benefits that optimized the existing Enterprise application stack.
- Campaign creation and management for this new class of customers is built out with queuing logic running on demand, against existing code and APIs. This approach eliminated the need to re-implement complex campaign logic for targets, tactics, advertisers, creatives, etc. Configured as a unified service, tenant context for each unique user/customer is managed via secure session tokens. Bids and placements from these small customers are thereby integrated into the same real-time analytics and bidding management engine.
- Historical reporting is largely reused from existing data infrastructure, to ensure data consistency. However, in order to speed this up for the new Pooled Services SaaS model. CloudGeometry also recommended refactoring the logic for existing enterprise-grade reporting services, to split it between a set of proxy servers and a back end. This reduced the reporting code base several fold, and cut reporting cycle time 10x across all users, both for SMB and enterprise.
- Moving from a small number of large customers to a large number of small customers required a new mechanism for managing payments. Cloud Geometry integrated payment card processing through a billing module that works with a commercial payment service. It records and reports the ad spending balance to each individual user/customer in their individual account
Key Benefits
By expanding to a Self Serve multi-tenancy platform, CloudGeometry helped this client reach new market tiers by extending their existing AWS enterprise platform capabilities. Without losing critical momentum of their high-growth enterprise core product team roadmap, the client could now unlock previously unavailable revenue streams. Its new, smaller customers can also take complete ownership of their advertising spend: to quickly create and manage campaigns; get real-time performance analytics; and achieve a virtuous cycle of ad spending ROI.
Tenant onboarding/ provisioning
Completely automated infrastructure resource provisioning via autoscaling of Amazon EKS pods to extend across existing AWS platform services
Secure tenant config & metadata
Per tenant configuration and metadata is stored securely in shared S3 and Amazon Aurora RDS DBMS storage; data isolation implemented on the application level
Data Science Automation
Streaming data via Kinesis tracks queues, async data processing in real-time. Cloudwatch, Grafana integration configured to specific internal platform APIs
