Case Study: Zefr achieves 6.6x higher video processing throughput and improved data quality with Appen

Brands now provide their contextual preferences which are stored and amplified using the Appen platform to fuel Zefr’s model for each client

Zefr, a provider of contextual video advertising that matches brand preferences to relevant video content, faced a scaling and quality challenge as demand grew. Their internal crowdsourcing program reviewed roughly 30,000 videos over two months (about 15,000/month) but lacked robust quality control and enough reviewers, forcing manual re‑review and slowing delivery—so Zefr engaged Appen and its crowdsourcing platform to find a cost‑effective, flexible solution.

Appen implemented a scalable crowdsourcing workflow integrated with Zefr’s moderators and context DMP, using Appen reviewers and machine learning to label and amplify brand preferences for each client. The result: guaranteed throughput and higher data quality, enabling Zefr to process about 100,000 videos per month (≈6.6× increase), eliminate much of the manual re‑review, and give customers precise turnaround times and quantitative quality metrics.

Zefr

Jon Morra

Chief Data Scientist