Another key to making RTB “Real-Time” is how quickly audience targeting can be updated. In our view, the old way of rebuilding a cookie pool every time an advertiser wants to tweak an audience segment takes too much time and results in wasted ad spend.  To fix this problem, we built an infrastructure that promotes targeting and bidding at a “data element” level, and allows advertisers to turn on and off targeting elements, or just change bids on individual elements instantly…with the new targeting generally live in about 5 minutes. This provides both transparency into audience segments, and the ability to optimize them automatically on the fly.

To use an example from search retargeting, if an advertiser sees that impressions served to users who have previously searched on the keyword “car rentals” are performing better than those served to users who searched on “cheap rental”, then the term “cheap rental” can be eliminated instantly from the audience targeting and the budget reallocated to more effective terms.

But I don’t understand why people are using any intermediary or replication steps at all.  This seems horribly broken to me.

At tickle.me we actually directly inject changes straight into what is referred to here as the “traffic” structure … instantly.  Our platform directly changes bidder behavior after any change is made (with the exception of content approval blocks), so there is no intermediary steps to propagate anyway and the design scales.

I think Google realises how influential it is in the market and that prophecies can be self-fulfilling.