Decoding the open internet: Supply path optimization

This article introduces a three-part series on lesser-known but critical topics for in-app advertising on the open internet, focusing first on supply path optimization (SPO).
Unlike walled gardens like Google and Meta, the open internet consists of millions of independent apps interconnected through mediation platforms, exchanges, ad networks, and DSPs.
A single ad impression can generate up to 40 bid requests from multiple exchanges, leading to bid duplication and marketers inadvertently bidding against themselves.
Inventory is non-exclusive—the top six exchanges cover 70% of impressions but only 1% is unique—so success comes from strategically accessing inventory rather than owning premium slots.
First-price auctions dominate, requiring precise prediction of user value and optimal bid price.
Machine learning processes 600 billion daily bid requests within 12 milliseconds, autonomously selecting the best path based on user valuation, win price prediction, and creative rendering features.
The key takeaways: the open internet is an incremental channel, there is no exclusive inventory, and advanced ML is essential for SPO to maximize ROI.