Challenge
Our client wanted to help advertisers understand what makes a social ad perform. To do this, they needed to ingest and normalize campaign performance data from multiple social platforms, spanning metrics such as impressions, link clicks and video completion rates, and connect it with the creative attributes of ad videos. This was a greenfield project with no existing tools or processes in place. The challenge was to establish a robust data foundation that could uncover performance drivers and provide actionable insights for optimizing future campaigns.
Solution
We started with a lean MVP to quickly demonstrate value and validate the approach with advertisers. Working in short iterations with frequent demos created a tight feedback loop, allowing us to refine features as requirements evolved and ensure early buy-in from stakeholders.
The AWS-based pipeline ingests campaign performance data into S3, cleans and transforms it with Lambda, and enriches it using Glue with PySpark to link metrics like landing page views, video percentile views and purchases to creative attributes such as brands and emotional tone extracted from each video. Data is then fed back into S3 to power the client’s SaaS dashboards, with Athena enabling ad hoc queries and anomaly checks. This architecture not only handles adtech challenges like incomplete data, shifting APIs, and changing schemas but also provides the flexibility to scale across new media channels.
Results
The client can now provide advertisers with dynamic dashboards that highlight which creative and campaign elements drive stronger engagement and ROI. These insights guide advertisers in optimizing creative strategy, audience targeting, and spend allocation.
The solution also established a scalable foundation for advanced adtech use cases. PySpark jobs can easily be adapted to push enriched data into a warehouse like AWS Redshift, enabling cross-platform attribution modeling, lookalike audience discovery, and predictive campaign optimization. Leveraging our adtech expertise, we delivered a production-ready MVP that meets immediate needs while paving the way for future innovation.