Pay-Per Transaction and Programmatic Voice

Recent remarks by Google executives have indicated that digital advertising is going through yet another major transition with the rise of internet of things, chat bots, and voice-controlled devices.

In an interview with Credit Suisse analyst Stephen Ru, Google's SVP of advertising and commerce, Sridhar Ramaswamy, had this to say, "We are keeping an open mind about the kind of monetization opportunity… It can range from being purely transactional — meaning we make it convenient for you to fulfill a transaction with this assistance — or it can involve promotion. But one thing that we are all clear about is the days of three top text ads followed by 10 organic results is a thing of the past in the voice driven world." [Emphasis ours.]

What does this mean for advertisers? In a non-visual UI world, how do they get paid?

Well it wound like the new kid (or business model) in town is the pay-per transaction model. In this model the user asks Amazon Alexa or Apple Siri to order something, the item is ordered, and a transaction occurs. If attribution can be made to an advertiser, a small portion of the transaction is carved out and sent to them.

But this is a dramatic change in the way things are done right now:

  • It requires very nuanced insertion so as not to destroy user experience. If I ask Alexa to play Gwen Stefani songs, I want to hear them immediately, so the opportunity to insert a quick ad spot is unlikely to work. However, if I ask Alexa what day it is and Alexa tells me, then also reminds me its my mother-in-law's birthday and that it can quickly place a same-day order from 1-800-Flowers....then I might welcome the interruption.
  • It requires cooperation from three different parties: the device maker (Google, Amazon, Apple etc.), the Advertiser (in this case 1-800-Flowers), and the payment processor (Visa, Master Card, PayPal etc.). This doesn't seem that challenging to make happen, the rise of Apple Pay, PayPal, Stripe and others has already effectively turn cash into an API-ready platform.
  • It requires better AI. In order for programmatic voice ads to work the voice recognitions have to get better at understanding something that has plagued artificial intelligence engineers for years: context. If I say "Alexa, order Cadbury Eggs for Chase and send them to her office in Alexandria." The programmatic ad system would have to understand that I want candy eggs and not real eggs, that Chase in this case is a person and not a bank, and that Alexandria is a reference to a location and not a person. While Alexa and Siri are getting better at this, they aren't good enough that they don't require correction. If the add is tied to the voice command, then corrections will really mess with ad attribution.

One thing's for sure, this will continue to be an interesting space to watch!