When you’re shopping for an iPhone on eBay, you might see an ad for a mobile service deal to complement your purchase. Or maybe you’re browsing through eBay Motors section and you come across an ad for car insurance. This is because when you shop on eBay, there’s a lot more going on behind the scenes, with advertisers competing in real-time for a chance to get your attention and show you something else that you might be interested in.
Let’s take that iPhone example: As soon as you hit “Search,” two distinct events are set off.
- eBay’s search engine looks for the best match to your search out of the 1.1 billion listings in our marketplace.
- eBay’s advertising exchange offers millions of advertisers the opportunity to show you an ad -- in real time! So, when you are deciding between the rose gold iPhone or the limited edition red one, various advertisers are competing for a chance to get your attention and show you their ad.
This advertising process is called programmatic Real-Time-Bidding. It is an on-the-spot auction, and operates similar to a stock exchange, with publishers and advertisers communicating through an exchange powered by machines.
On nearly every page load, eBay runs these types of real-time auctions for ad impressions. But, the standard way of programmatic advertising can slow down webpages and distract buyers from their shopping journey.
We want you to get the best search results for that iPhone -- or whatever you are looking for -- as fast as possible, without letting ads slow down the process or disturb your shopping journey. Using programmatic science technology, we built an innovative, unified auction exchange called “parallel bidding.” By building our own advertising technology that leverages server-to-server programmatic pipes, we can improve advertising yield, reduce advertising latency, and enable data-driven control of advertising costs.
In a process that only takes a few hundred milliseconds, parallel bidding sends out bid requests to multiple advertising bidders in the exchange in a parallel, simultaneous fashion and replaces the older and slower method of programmatic advertising that sends requests to bidders one at a time. This allows for a truly fair and unified auction, where all advertising bidders have an equal shot at winning the impression and drives up yield. This bidding process also collects bids for ad impressions behind the scenes on the server-side, speeding up the time it takes to load an eBay page and ads on your browser, and improving ad latency by up to 60%. And parallel bidding is in fact the auction house for our programmatic advertising, running and clearing an auction for each ad impression on eBay. The data it collects to make these auctions possible, coupled with eBay’s internal user behavior data, allows for intelligent pricing and cost control.
Our goal is to create a valuable experience for both buyers and sellers, while also driving performance for advertisers. We strive to bring together two marketplaces: the shopping marketplace, connecting buyers and sellers; and the advertising marketplace, connecting advertisers and customers.
We are continuing to work on parallel bidding with the goal of growing it so that it becomes the main way of showing ads on eBay, benefiting our entire marketplace.
Amir Green is a Senior Product Manager at eBay in New York City. He is product owner for the programmatic science team at eBay Advertising, focusing on unified auction mechanics, and building innovative technological solutions that bridge e-commerce and advertising technology.