Real-time bidding (RTB for short) is all the rage in modern media buying. Sophisticated technology companies like TellApart and DataXu claim (and probably actually can) analyze an available impression on an ad exchange like RightMedia in under 120 microseconds and determine whether it is worth bidding on. The data these systems look at include the placement, time of day, geographic, ad unit and many, many more. Indeed, DataXu claims that they can instantly process more than 100 data points, as seen by the graph here:
For data geeks like me, the idea of being able to evaluate each search query instantly and make a precise bid based on reams of data is sort of the Holy Grail. PPC systems, however, don’t come close to this level of granularity. The vast majority (if not all) campaign management software packages optimize bids once a day. Geo-targeting and day-parting are available, but they are pre-set, not real-time evaluations of data.
You might think that creating a system that literally bids for each query individually and in real-time would be impossible. In actuality, it already exists. It’s called the Google Conversion Optimizer. Indeed, when you read the description of the Conversion Optimizer it sounds an awful lot like what you might read in a ‘How it Works’ section from a demand-side platform (DSP) like TellApart, which, in the case of TellApart, was build by ex-AdWords product managers and engineers:
So why not make this data and RTB functionality available to SEMs? Surely campaign management companies like Marin Software would love to build algorithms to handle RTB; indeed, I suspect that RTB algorithms would create tremendous bidding efficiencies (and enterprise value) for campaign management software.
One answer may be bandwidth. The number of players in media buying is infinitely smaller than the millions of PPC advertisers. To participate on an ad exchange like RightMedia or the DoubleClick Ad Exchange, you generally need to buy a “seat” much in the same way that brokerages buy seats on stock exchanges. These seats can be expensive and serve to reduce the number of active bidders on the exchange. Imagine if every search agency, campaign management software company, and large and savvy advertiser wanted to participate in RTB on AdWords. Instead of one or two API calls a day via the AdWords API, each advertiser might make tens of thousands of calls. That’s a lot of servers, or HADOOP, or something (trying to sound smart here).
Another answer may be demand. It may be the case that no one in the SEM industry has bothered to ask about RTB. This may be due to ignorance (not knowing it is available) or simply that the current “bid once a day and set your geo’s” system is ‘good enough’ for most advertisers. Perhaps as display and search converge, more advertisers will see what is going on in display and say ‘hey I want that for PPC too.’
A final reason may be that Google is just unwilling to provide the data. There’s a real financial incentive to provide more data granularity in display; simply put, without RTB and deep data, display media will be forever the domain of big, inefficient brands. Providing better data enables Google and other publishers to attract performance marketers with deep pockets, assuming the CPAs work out for them. It also enables publishers to create a bidding environment for inventory that would have previously been sold at below-remnant prices to middlemen ad networks. So RTB results in more buyers and more bidders and more money for publishers (at least quality publishers).
RTB for PPC may not have the same positive revenue impact. You could argue that providing some but not all available data gives PPC advertisers a sense of control and granularity but still preserves some healthy inefficiency (and therefore margin) for Google. Why provide more data that may or may not benefit Google?
I don’t know the answer, and probably anyone who does doesn’t read this blog. Online marketing is about data and really SEM was much faster to embrace data than display. It’s ironic that the tables have turned so quickly.