As the difference between winning an ad request dwindles down to microseconds, advertisers are beginning to distribute their servers geographically so they can respond as quickly as possible. It is not uncommon for a company to have servers located in different regions (ie US West Coast, US East Coast, Europe, etc.). This allows the ad request to be sent to the nearest ad server location in order to speed up the response time. If the ad request originates from the East Coast, then sending it to the Eastern server instead of the Western one can make the difference in ensuring that the ad response is returned within the time limit.
The increase in response time can be partially attributed to user data matching. As advertisers become more granular in their user targeting, they are matching the user data passed in the request (device ID, cookie/User ID, etc.) against their internal data store or third party data. The matching involves identifying previous user performance (click through rate, view through rate, conversion rate, etc.) and any additional demographic data (gender, age, etc.) available about the user. While this additional data helps the advertiser to determine if they want to show an ad to the user, it also comes with a cost, namely time. The time required to do this data look up, which can be even greater if going out to a third party, can inhibit an advertiser from responding within the time limit.
Ultimately, distributed servers and greater data stores should improve response times and CPMs for publishers. The distributed servers should decrease the response time since they should now be closer to publishers and the more granular targeting capabilities should increase the CPMs, but that has not been the case. Response times have have gone up as advertisers try to do as much analysis as possible and while CPMs have risen for some premium publishers, the additional data matching has become the norm and has resulted in a drop in CPMs for requests that do not have all of the required data to do the matching.
Increased advertiser response times can have a profound effect on ad servers as they must now wait longer for a response or until the timeout limit is reached. In addition to slower responses, ad serving companies have incurred additional costs from distributed servers and additional hardware to try and maintain as quick a response as possible for the user. This additional time translates into a poorer experience as the user must wait longer for the content to load. Eventually response times should get quicker as technology advances, but in the meantime advertisers need to be sensitive to the impact their data look ups and slower response times are having on the ad server and ultimately the user who will be viewing the ad.