With the explosion of data (which is now cheaper than ever to store and to process), predictive analytics has become a hot and growing field. The data is being put to uses as diverse as forecasting traffic, recommending movies, predicting clicks, preventing fraud, and even classifying a fading pitcher.
It was on the topic of predictive analytics that MIT/Stanford Venture Lab (VLAB) held a very well-attended and insighful panel discussion, featuring Omar Tawakol of BlueKai, Scott Burke of Yahoo, Matthew Barkoff of Badgeville, and Theresia Gouw Ranzetta of Accel Partners, with Michael Driscoll of Metamarkets serving as moderator.
Without attempting to provide a word-for-word transcript of the presentation, let me cover a few of the interesting points that stuck with me as well as my own thoughts/interpretation.
History
Before predictive analytics, how did advertising companies know who to target? They used proxies. A zip code for an affluent neighborhood may serve as a proxy for the kind of discretionary income that the homeowners have in that neighborhood. A TV show might be a proxy for the target viewer of that TV show. So companies would sell (and still do) direct mailing packages into specific zip codes based on the demographics of those zip codes, or offer advertising time during TV shows, as a few examples.
But how much do you really have in common with your neighbors in terms of the kinds of products and services that you are interested in? Or with the other viewers of your favorite TV show? Not very much I would guess. Proxies are obviously a very imperfect method of directed marketing.
Cue: enter predictive analytics to save the day. :)
Data Collection
Predictive analytics obviously needs data to feed its algorithms. So where does that data come from?
A portion of the data is collected by online service providers.
When you search on Yahoo or Google or Bing for designer shoes, baby cribs, corporate gifts, European cruises, or wedding photographers, the company providing the search engine capability aggregates that data to know what your interests, hobbies, and buying patterns are in order to place more relevant and therefore more expensive advertising in front of you while you search.
Amazon, formerly a large online book store and now a supermart for everything under the sun, from video-on-demand to Halloween costumes to high-end electronics, amasses a huge amount of data over time about the shopping habits and preferences of its frequent shoppers. [Though to be honest, I have not seen them put the data to good use with their shopping suggestions, which never seem to inspire, just mho.]
And finally, social networks like Facebook or Yelp know as much as you are willing to share about everything in your life from where you went on vacation, to your favorite restaurants, to the kinds of networking events that you attend.
The common trend in these three examples: the data is provided by the users, whether consciously or not.
And I, for one, don't see anything nefarious in this. [Gasp!]
Rather, I think of it as a kind of implicit pay-to-play. Only the currency to play isn't cash; it's personal information.
Which leads us to the next point.
Data Ownership
Following the logic from above, if the price for using free services like Facebook, Yelp, or Google search is the personal data collected while using the services, then ownership of the data rightfully transfers to the service provider at the time of use.
However, because the pay-to-play is only implicit and because of the... well... rather personal nature of personal information, the transfer is probably more akin to a limited license than a full assignment, to throw around some legal jargon.
[Please note that the theoretical legal framework I am suggesting is artificial and fictional, and is only meant to faciliate discussion, not provide a legal opinion on the underlying transactions. :) ]
Data Use
If, following the premise above, the license is indeed limited, the big question is, what can the companies that collect the data do with it?
And I think the answer is, to the extent that the data is anonymized and handled properly and securely, they can sell it to advertisers/advertising agencies (via data marketplaces like BlueKai, for example). The caveat for data being "anonymized" and "handled properly" and "securely" is supposed to protect against broad typoes of misuse of the information, including identity theft.
But short of that (misuse that is), how bad is it really if the marketing aimed at us is (even incrementally) more relevant to our interests than the same amount of marketing that is instead total junk?
Your Options
Not every website with let you do this, but a growing number of websites, Yahoo, Google and Bing among them, will let you manage how that website perceives your interests (and, therefore, the kind of advertising you might expect to see).
If you are interested, at the very bottom of the Yahoo home page, click on "About Our Ads" and then proceed to the gray "Manage" button to see the kind of categories Yahoo is using to assess relevance in placing ads on your pages. Very similar process on Bing. For Google, click on "Privacy" at the bottom of the page, then on "Ads Preferences Manager."
The Future
I don't have a crystal ball, but I would not at all be surprised to see more companies allowing greater transparency and user customization of interests used as a basis for interest-based advertising. Maybe that's the next evolutionary step in advertising, in fact, following proxies and black-box predictive analytics.
Inna Efimchik
 | Emergence Law Group, specializing in assisting emerging technology companies in Silicon Valley and beyond, provides incorporation, financing, and licensing services as well as general corporate counseling. |
