This is our first customer newsletter and we think it is important to address the elephant in the room right up front. From the time we introduced our VPAID Ad Manager in early January through late April, we had significant discrepancies with 3rd party demand sources in our reporting. We fully recognize that this made it very difficult for our customers to be successful using AdForge when using video ad serving.
In April, after our previous efforts to solve this problem had failed, we took a different approach and literally searched the globe for experienced VPAID & Flash developers to supplement our own team’s capabilities. We ultimately assembled a war room and flew both employees and new outside contracts into San Francisco where we spent 9 12+ hour days assembled in a room until we solved every last discrepancy with 3rd party demand sources. In the end we were successful and the silver lining in all this is that we went from being highly discrepant to almost perfectly aligned with every other ad server in the market as the chart below indicates.
Although we have to find out why there are some large discrepancies with two of the demand sources.
Ad Serving Latency Chart
We recently made several code changes that have reduced server latency by 38%. Below is time series showing average server latency across our server clusters. 29.6 ms is very fast. But as you walk through the three charts below you will see that the go from 29.6 ms to 25.9 ms and finally down to 18.1 ms. There are two reasons for this big drop in latency. First, we introduced query caching and in the 2nd chart you can literally see where SQL calls disappear as a time factor. Following the introduction of query caching, we then performed a number of other code tweaks which in the end have dramatically reduced overall server latency. While there are many factors that affect fill, anecdotally it looks like we are seeing an 8% increase in fill as a result of this server latency reduction.
AdForge Admanagers “Don’t sleep”