Cache-busting with GPT
When GPT assembles ad requests corresponding to a page view, it automatically includes a "correlator" value. This value is common to all ad requests for that page view, but unique across page views. This correlator mechanism provides two key benefits:
- It serves as a cache-busting mechanism, removing the need for a developer to append random number parameters.
- It allows the ad server to identify when ad requests belong to the same page view, which enables roadblocking and competitive exclusions, and prevents duplicate ads from serving to the page.
Our ad servers maintain this correlator value briefly (currently for 30 seconds, but subject to change), such that requests with the same correlator received close together will be considered a single page view. If subsequent requests come later, the page view will be considered new regardless of the correlator value. An expired correlator will impact any feature where the creatives or line items eligible for the current slot depend on the line items or creatives shown in previous slots, such as competitive exclusion and roadblocking.