Case Study: Major Media Company boosts ad segment performance with Vidora (mParticle Cortex)

How many positive samples you need to build an ad segment

Major Media Company worked with Vidora to improve how it builds premium ad segments for advertising monetization. Their challenge was determining how much first-party declared data was needed to create high-quality look-alike audience segments, since collecting that seed data can be costly and time-consuming. Vidora’s Cortex product was used for the analysis.

Vidora used Cortex to upload seed data and train look-alike models across five different seed set sizes, repeating the experiment 10 times for statistical significance. The results showed that larger seed sets improved performance, with the case study reporting up to a 5x increase in segment performance and an average rank score above random even with as few as 10 positive labels. Overall, Vidora helped Major Media Company confirm that more seed data meaningfully boosts look-alike model quality.