In this guide

A bidder returns attractive prices, but many responses arrive late. Increasing the timeout seems like an obvious way to recover the revenue.

That assumes the extra bids are still useful when they arrive and that waiting has no cost elsewhere. Both assumptions need testing.

In header bidding, several demand sources can compete before the ad-serving decision. The auction’s timing determines which responses can participate and when the page proceeds.

Identify which timer you are changing

A setup may include a bidder timeout, a wrapper timeout, an ad-server request schedule, and additional delays before the auction starts. They are not necessarily the same timer.

Prebid’s configuration reference documents timeout behavior and related settings. Inspect the installed version and implementation rather than copying a value from an unrelated site.

Record when the auction begins, when bids arrive, when the ad server is called, and when the creative renders. A delay before the auction will not be fixed by shortening the bidder-response allowance.

Look at the distribution

An average response time can hide a long tail. Review how many eligible responses arrive within different intervals, segmented by device, geography, connection, and bidder where possible.

Suppose 90% arrive within 700 milliseconds, another 5% by 1,000, and the remainder later. Those illustrative figures describe response availability, not the revenue benefit of waiting.

The late responses still need to be eligible, win, and render. Count the value that survives those stages rather than summing the prices of every late bid.

Test the whole page outcome

Compare a controlled timing change against the existing setup over equivalent inventory. Watch auction participation, rendered impressions, revenue per comparable page or session, and user experience.

A higher bid CPM can coexist with fewer rendered ads if visitors leave before loading completes. Conversely, a slightly longer wait may be worthwhile in a particular environment. The evidence should decide.

Avoid changing bidder configuration, floor rules, and timeout simultaneously unless the test is explicitly designed to evaluate the combined setup.

For pricing changes, the floor-price guide explains another way a winning CPM can misrepresent the overall result.

Check failure and retry behavior

Review what happens when bidders fail to respond, consent prevents an auction, or a page component initializes twice. A timeout should not leave the ad slot permanently blocked or cause unintended duplicate requests.

Prebid’s requestBids reference describes the auction request interface. Developers should verify the behavior applicable to the deployed version.

Save the tested configuration and a reversal path. The best timeout is an observed tradeoff for the site’s traffic and implementation, not a universal number.

A useful decision record shows which extra responses became useful delivery and what happened to the page while it waited.

AdOpsNow Editorial

Examples use illustrative data. See our editorial standards or report a correction.