Why does header bidding cause latency? How can we reduce it?
Having seen almost all bidders’ header bidding API calls, we’ve observed the few problems listed below:
The first two scenarios show websites’ network calls after implementing header bidding.
The final scenario shows how header bidding is accelerated by Prebid.js.
All Pre-bid Calls are made concurrently within 100ms: Note that AppNexus, Pubmatic, OpenX, Rubicon bidding header calls were all made within the first 100ms.
The timeout of 400ms is respected: We set the timeout to 400ms. As you can see from the graph, the GPT tag (gpt.js
) is loaded at around 500ms. The reason that GPT didn’t get loaded exactly at 400ms is that the JavaScript timer is nondeterministic. Some partners take longer than others. The ones that took longer than the timeout setting did not get a chance to bid due to latency concerns.
Rotate order of bidders: To help publishers maximize yield, all header bidders should be given the same amount of time to respond. However, JavaScript doesn’t make calls exactly at the same time, so we help you rotate the order in which the bidders are called.