Reproduce “Audience Prospecting for Dynamic-Product-Ads in Native Advertising” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Eliran Abutbul, Yohay Kaplan, Naama Krasne, Oren Somekh, Or David, Omer Duvdevany, Evgeny Segal

Yahoo Research and Tech Yahoo teams developed two new audience prospecting methods, Conversion-Prospecting and Trending-Prospecting, for Dynamic-Product-Ads (DPA) within the Yahoo Gemini native advertising marketplace. These approaches, leveraging the proprietary OFFSET algorithm, increased DPA delivery by 10-13% and revenue by 1.58-6.33% in A/B tests, while maintaining strict performance targets, and now comprise over 80% of DPA prospecting traffic.

Key insights

  • Integrating user features like "ctr-campaign-top" (past click behavior), "age", and "gender" substantially improves the accuracy of DPA prospecting models.
  • Dynamically calculating a target Cost-Per-Action (CPA) for new users enables conversion-focused prospecting to expand audience while adhering to advertiser performance goals.
  • External pixel feeds can be effectively leveraged to build lookalike audiences for new advertisers or products, addressing cold-start challenges.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.