Computational advertising – where are we?

Computational advertising has grown by leaps and bounds over the past decade. As Digital marketing techniques and platforms grow, there are  a number of new open problems that are on the horizon spanning a range of fields. I briefly summarize them here.

  • Audience Behavior – How much more do we know about the “audience” in  a reliable manner before the days of online advertising? Can we know more without violating the bounds of privacy?
  • Content Analysis – How do we integrate content analysis across text, audio, video, images to get a reliable view of the audience and brand perception?
  • Integrated Channel Management – How do we coordinate campaigns across email, web/SEO, SEM/PPC, social, mobile, video, audio, radio (online/offline), print and TV?
  • Active vs Passive Engagement – Transition from a one-way mass communication paradigm to a 2-way – one-on-one conversation puts immense load on both the marketing infrastructure and the individual audience member. Is this the way forward? or is there a happy medium?

With all the technologies in play, the advertising “workflow” is still open-loop. Marketers message people (in a number of ways – push or pull), and the audience responds either online or offline, near instantaneously or in a delayed manner co-located with the message or not. All the data available through this workflow is noisy and fragmented either by design or by operation. However, the connection between audience response and the initial marketing strategy is made in a heuristic manner. Filling in this gap in a reasonable manner would be a worthy challenge for different players in the eco-system.