At the Pervasive 2009 conference in Nara, Japan, I co-organized a workshop on Pervasive Advertising which examined the impact of pervasive technology on the future of marketing.
Background: the place & the workshop
Nara is a spectacular place with pristine, exotic Buddhist temples and some of the friendliest — some say “holy” — deer on the planet. (The deer are really very cute.) I stayed in the smallest hotel room ever – just barely avoided ducking into the bathroom because the floor was raised about 6 inches high and the ceiling was about 2 feet low. It wasn’t bad though; there was actually enough room for one medium-sized person like me.
The quality of the workshop talks and the people was quite high and I expect this will be a community with which we continue to collaborate:
- The majority of the organizational effort was really done by Jörg Mueller. The other co-organizers were professors of ubiquitous computing Albrecht Schmidt and Aaron Quigley.
- I gave a talk on Activity-Based Advertising: Techniques and Challenges, which was a position paper that Kurt Partridge and I wrote together (Kurt did most of the work!)
- I also gave an unplanned presentation on Responsive Media which involved collaborations with our colleagues in ethnography (James Glasnapp, Brinda Dalal) as well as computer vision and sensing (Wei Zhang, Maurice Chu, Julia Liu).
PARC’s approach to responsive media
There are many researchers looking at technologies that can detect a human response using cameras and other sensors that pick up demographic data (gender, race, age) and physiological states (eye gaze, orientation, pupil dilation, skin temp, expression, etc.). There are a variety of applications that are envisioned for these technologies including human-robot interaction, marketing, gaming, digital concierge avatars, etc. We refer to this class of applications as “Responsive Media” and it is one of the most exciting areas of current research in human-computer interaction.
At PARC, we take a strongly human-centered approach to inventing Responsive Media technologies. Whereas most other research labs tend toward a technology approach (e.g., what can we detect using a particular technology such as visual or thermal spectrum cameras or other sensors), we focus more on identifying significant human behaviors using ethnographic/anthropological methods – and then we invent a technology to detect the behaviors. At the workshop, for example, I showed video clips that our researchers had collected showing that certain human poses indicate non-commitment to engagement. Another video shows behaviors that indicate when a person has finished speaking (hands at “home” position, eye gaze flick, etc.).
Another example of our strongly human-centered approach includes detecting clothing characteristics instead of simple demographics (gender, age, ethnicity) which has been the focus of other research labs. In our modern society, what you wear says more about your taste than your demographics. You can’t do much to change your gender, ethnicity or age, but you do choose what to wear. Marketers prefer this psychographic profiling over simple demographics because it indicates the interests, activities, and opinions that a customer values.
An example scenario: Pervasive Marketing in 2024
After the paper presentations, we formed working groups to brainstorm about future scenarios of marketing technologies. One recurring idea was that of a person becoming a walking advertisement and broker for the products he is wearing or using (e.g., clothes, bike accessories, etc.). The sub-group I was in came up with this scenario to illustrate the concept.
- After work on a hot and humid day in the Summer of 2024, Sabine and Monica walk out to their gravity-powered vehicles.
- Monica shows Sabine that she just downloaded the Danica Patrick street performance tuning package for her GTO Nouveau and it sounds much cooler than the high-pitched woosh these gravity-powered vehicles normally emit! Sabine loves the “retro” sound of a combustion engine and exhaust and she gets a copy of the tuning package for her roadster.
- As a registered “Danica Patrick broker”, Monica’s e-money account gets credited with a portion of the proceeds from Sabine’s purchase. Monica can only use the funds in this account for vehicle-related products and Monica has to manage her multitude of brokering accounts in various product categories to get everything she needs. Fortunately, her digital accountant manages the complexity and lets Monica know when she needs to take some action.
- On the weekend, Sabine goes out with her family for a bike ride. At a rest stop, Sabine’s bicycle suggests she look at the bike she’s parked next to because it has tires that would fit Sabine’s riding style better than her current tires (Sabine has an off-road bicycle but actually rides most often on trails and roads, so street tires would be better). Sabine looks at the other tires and her system adds them to her online “wish list”. Later when Sabine buys those tires, the owner of the other bicycle gets a cut of the profit as a member of that tire company’s broker consortium.
Enabling Technologies: What makes this scenario possible
The brokering system is a complex, but already feasible account tracking system, so there’s not much needed to invent that part of the scenario. Beyond that, the underlying system keeps a record of Sabine’s activities and the items that she encounters during the day, which allows her to look back in the record of a day to see items she encountered that might be of value to her. Again, this may have scalability issues, but doesn’t require fundamental new inventions.
There is more invention needed to create the psychographic preference model, however. The system:
- Continually updates the model of Sabine’s preferences based on what items Sabine actually expresses interest in
- Understands that Sabine sometimes shops for other people (her kids and gifts for friends and family)
- Detects when it has been more or less successful in timing the presentation of information, determining whether to tell Sabine about products at times when she is likely to be more interested in the information (such as after the tennis game, not during a serve)
- Detects Sabine’s points of decision — that is, the moments in time when she is considering whether she will or will not make a purchase
In cases where the preference model does not already have information that a seller would want to know about, the advertisers can submit queries to her model which process the information without exposing Sabine’s personal info to unwanted scrutiny. For example, suppose the system was not monitoring Sabine’s bicycle riding style, but it does have records of all of her prior rides. An advertiser could then submit a query to her model that would execute a function to classify her riding style without exposing all the location details of her prior rides.
Workshop action items – your input would be appreciated!
We decided to produce a short book of expanded versions of the workshop papers targeting a broad audience — not focused on the details of technologies but on the potential for new applications and business. (Jörg will probably be coordinating that.)
When we run the workshop again, we’re going to invite participation from marketing or product groups to get more from the business community and hopefully hold the workshop in conjunction with a business/marketing research conference. We need to bring the technology and marketing communities together so we don’t end up with Pervasive Spam or worse!
So, dear readers, please suggest marketing/business innovation researchers with an interest in co-organizing workshops on these problems?
Editor: Sonal Chokshi