I originally wrote the entry as an online article and it ended up being quite long. So I decided to start with an extra summary to present the core idea briefly. Well, it made the whole thing that much longer.
Advertising connects demand with supply: it connects people who have money to spend to fill a need (or to splurge out on a luxury) with businesses which want to reach customers. It is in the buyers’ interest that they are only exposed to advertisements that help them in their purchasing decisions. It is in the sellers’ interest that their adverts mainly target consumers who are potential buyers of their goods or services.
In online advertising, buyers need information about products and sellers, while sellers need help to target their ads at potential buyers. Help might come from a dedicated provider (either by people employed for this task or in an automated way from some remote server), while the alternative is that it might be provided by some online community. These are independent aspects, as the table shows with some examples.
|Recommending…||seller to buyer||buyer to seller|
|Centrally/algorithmically||independent product reviews; Google AdWords, Amazon, eBay||browsing history (Google, Amazon)|
|In a distributed fashion/socially||Amazon, eBay; Pinterest||?|
My observation is that I do not know any participant in the field with ? but Facebook is in a position to enter that space.
What concerns Facebook’s near-term financial health, the desktop versus mobile debate might be a misplaced one. If Facebook is to monetise its user base, it could encourage its users to recommend their friends to advertisers.
Eric Jackson in his musings about the future of today’s Internet giants in the face of the mobile internet revolution, and Jay Jamison’s inventory of challenges facing mobile internet revolutionaries-to-be identify three generations of internet companies. Web 1.0 companies were founded from 1994 to 2001, and include Yahoo!, AOL, Google, Amazon and eBay. Web 2.0 or social companies predominantly came about from 2002 to 2009, such as Facebook, LinkedIn and Groupon. The latest breed are the mobile internet companies starting from 2010, like Instagram.
The lofty valuation of Facebook is based on the hope that it can pull off the stunt that Google did and can monetise user data through customised advertising. By building on the enormous amount of data it collects, it tries to match users to banner ads which they will click on. The recommender systems of the likes of Amazon, Netflix or Last.fm are not very different conceptually, either: feeding data from logs of other users’ interactions with the respective services into an information filtering algorithm, goods, films or songs are recommended to the targeted user that will likely be met with interest.
Data mining used in these systems is not specifically an internet technology by any means, still, if we try to fit it into the timeline, it was already used for Web 1.0. However, the social aspect, namely, user ratings and reviews of products, has become a fundamental feature of the service that, for example, Amazon provides. As opposed to having an answer ready for the user, Web 2.0 creates environments where other users can provide the answers.
Matching seller and buyer is a non-trivial service that, if done well, deserves to draw a fee from both sides, according to Gregory La Blanc. Online shoppers on the one hand rely on product profiles (reviews, ratings and recommendations) to make their purchasing decisions. Marketers and retailers on the other hand use profiles of internet users to advertise their products to, profiles that arise from internet searches and browsing history. Both the buy and the sell sides have options to choose from, hence each side ranks their potential counterparties based on predicted utility or profitability.
Products can be rated through objective tests: there are numerous product-testing organisations and publications. Products can also be rated socially, that is how Amazon’s user-generated reviews work. They can even be rated by friends via sharing on, say, Pinterest. Users can be rated algorithmically, that is how Google and Amazon do it. But Facebook is in the sweet spot to be able to rate customers for the advertisers socially. Facebook could build the means to let users rate and categorise their friends for targeted advertising. That would be a social, user-generated type solution to a problem where mainly technological solutions have been sought so far. (Bear in mind, of all things, Google Maps is also heavily reliant on human insight, although there the employees are just as important as users.)
Facebook could incentivise its users to opt in to participate in labelling other users by one of several approaches. First, Facebook could split the income the clickthroughs generate with the users whose suggestions led to the display of the successful ad. Second, Facebook could offer credits which could be used inside Facebook to purchase extra services. Third, Facebook could formulate this task simply as a fun activity, as short surveys at login or at random times.
In order to prevent collusion between users in the case of tangible compensation (you tell Facebook I have some niche interest which I do not, I always click on those ads, and we divide your earnings), a number of measures seem necessary. The targeted user must not know who recommended her. It is helpful if a recommendation is acted upon only if multiple users independently made it. And it can be counterproductive to let participants choose whom they recommend for better targeted advertising, so some randomisation of whom the participants can recommend might be required. In contrast, in the case of the third approach, revealing among friends what was said about one another might be part of the fun.
One can imagine a Facebook application randomly giving names of friends, whom the participant needs to associate with categories of interests, similarly to the categories in Google Ads Preferences or otherwise (e.g. Tom — haute couture or nerdy? Dick — always the first to have or lives in the Stone Age? Harry — rafting or ballet?). Or, what seems more practical from the viewpoint of Facebook and its advertisers, the participant would be given the product categories and would need to recommend interested friends (which three of your friends tinker with their phones the most? whose advice do you trust when you need a new computer? who are the biggest petrolheads/fitness-buffs/football fans among your friends? boys and their toys — who springs to your mind?).
Facebook users know much more about their friends that help advertisement targeting than merely product group interests. Examples include approximate disposable incomes or inclinations with respect to the innovation adoption cycle: are they early adopters or more conservative types? Users might learn early that a friend is considering buying a big ticket item and can share this information with Facebook. (Here the risk of collusion requires further thought.)
But would Facebook really want to implement something like this? Some will argue that selling out friends in this way is cynical. Many among us object to the selling of their personal details to marketing agencies. The monetary incentivisation approach would certainly come up against a brick wall of moral and privacy concerns. To some, any arrangement of this type would evoke methods of secret police operations: friends reporting on friends to an intelligence gathering agency. Therefore such a move can be expected to alienate many users and to cause a public outcry.
The first method would also be dogged by the fact that very little money could be made; for most of us, clickthroughs could not possibly earn enough to be worth our time. After all, this is one of the reasons why we trust computers to pick display ads: their labour is cheap. With an approximate $1 cost per click for Facebook display ads currently, if Facebook redistributed half its additional earnings from the proposed scheme, then to achieve the US federal minimum wage of $7.25 per hour, one should aim to bring in 14.5 additional clickthroughs with an hour of work.
The obstacle to the second approach (credits as incentives) is that `It’s free and always will be.’ is one of the central principles of Facebook. Facebook would need a shift from free to freemium to create the preconditions for this approach.
The third one is the least problematic and minimises the risk of manipulation. LinkedIn already features a skill endorsement system in this manner. Facebook also has had a campaign where users are asked to identify if selected account names are real or pseudonyms. Here the challenge is to design an activity which engages users without offering a tangible return.
The arguments for this development would be that it is in line with Facebook’s taste for risk-taking and `Move fast and break things’-culture, with its Advertising Philosophy and has the potential to improve user experience. Facebook could become the place on the internet to hang out where not only the shared content, but even the display ads are so intriguing that they distract the user and tempt their senses to explore colourful, exciting websites of cool gadgets or online fashion and style retailers or whatever that tickles their brain.
Would this move make Facebook a better loved social networking service? Not necessarily. Would this make sense for Facebook financially? If they can limit the backlash of users, then probably yes.