Last week I blogged about a few research projects presented at the Wharton conference on user-generated content. In this second part of the conference summary series, I’d like to discuss one other interesting presentation that is not as directly related to user-generated content per se but I think can be of tremendous interest to online advertisers. Then to wrap up the series, I will list a few research questions raised by industry participants at the conference. This will probably be particularly interesting to researchers who are wondering what is on the practitioners’ mind. By the way, the conference has created a page with links to all the presentation slides.
http://www.flickr.com/photos/teofilo/ | CC BY 2.0
How to target ads to consumers without sacrificing their privacy?
The recent controversy surrounding Facebook’s privacy setting changes shows us that privacy issues are still very much on people’s mind these days, especially with a large amount of very personal data now available through online social networks. To advertisers, the increasing amount of social and personal information represents a great opportunity to offer very targeted ads to consumers. But as we get closer to consumers’ personal domain of interests and friend networks, advertisers are also treading a very dangerous water of consumer privacy. This is why I find New York University Professor Foster Provost’s research to be particularly interesting, as it allows target advertising toward consumers while still protecting their privacy, or in the researchers’ term, “privacy-friendly” target advertising.
The basic idea is quite simple, although the actual implementation can become more complex and mathematical. The underlying premise of the approach is that consumers who are more similar to each other are more likely to buy the same brands and share similar consumption habits. This is why social network information can be very powerful, because we are likely to buy the same things as our friends or at least have a good deal of influence on each other. The problem with using such explicit social network information is the privacy issue. To circumvent this problem, Professor Provost’s approach uses anonymized browsing data instead. It builds on two key sets of information: (1) a set of consumers who are considered brand actors; and (2) browsing data for these brand actors and other consumers whose brand affinity is not yet behaviorally demonstrated.
For the first set, one can use criterion such as having visited a brand’s website or fan page on Facebook to identify consumers who are brand actors. Notice that advertisers do not need to know who these consumers actually are in terms of names or demographics, but just that they are entities who have demonstrated certain desired behavior. Then with this information, the brand proximity/affinity of other consumers can be calculated by analyzing how closely the content (brand and non-brand related) visited by those consumers resemble that of the brand actors. Potential consumers can then be ranked based on this similarity to identify the ones that have the closest brand proximity. Professor Provost’s research shows that consumers picked in such a fashion have a much higher concentration of potential brand actors than random picking and that these consumers are much more likely to be linked to known brand actors. A paper from this research project is available from Professor Provost’s website.
To me, the beauty of this research is two-fold. First, because the only data needed are browsing logs without personally identifiable information attached, it allows advertisers to selectively target consumers without having to worry about privacy issues. Second, because the approach is defined in a sufficiently general fashion, it allows for much tweaking and customization. For instance, various brand proximity measures can be used (this research itself suggests five measures), and different measures can be combined to most accurately gauge brand affinity. Moreover, the criteria used to spot brand actors can be customized based on an advertiser’s needs (e.g., visit to awareness page vs. conversion page depending on the goal of the campaign). Such flexibility makes the approach applicable to a wide variety of situations.
What do practitioners want to know?
The conference organized a few industry panels to talk about their own experiences and their unanswered questions. Out of these industry participants, Mr. Gary Spangler, E-Marketing Manager from Dupont, spoke the most systematically about a set of research questions that need to be addressed from a practitioner’s standpoint. Many of these questions were echoed by other industry participants. I list them here for the benefit of academics who are in search of practically relevant research questions.
- There are more and more ways to reach/touch consumers. Is there a way to analyze the value of each electronic touch (e.g., email, social network, etc.)?
- When lead time is relatively long (e.g., 1 year or more in the case of B2B marketing), how does one measure the ROI of online marketing investment? (We all know that ROI has always been an issue, but longer lead time apparently posts an even greater challenge.)
- How can a company use information from web queries (similar to the browsing information used in Professor Provost’s research described above) to identify potential sales leads?
- When potential leads abound and resources available to respond to those leads are limited, can we develop a lead scoring system so that a company can properly filter out more important vs. less important leads?
- Different online marketing approaches use different types of content as input. For example, a company’s website and its social network presence most likely require different content. How can one measure the value of each content type to different segments and different industries?
- Demonstrate the ROI of social media efforts to help marketers argue the value of social media participation to upper-level managers.
In us academics’ constant quest for new knowledge, questions such as these are very useful in guiding our research effort toward being more relevant and applicable to practice. Here I send out a call to practitioners out there to supply us with more of these and to tell us the question marks in your head. Please feel free to leave your comment here. As the overarching goal for my blog, I would like to make Ping! an intersecting spot for practitioners and academic researchers.
This is going to be my last blog before Christmas. So here’s happy holidays to all my readers. Wish everyone a warm, safe and love-filled holiday!
