The End and Limits of the Facebook Model
I’m tired of reading that Facebook is going to change. I’m worn out about hearing how important privacy is to most companies, while they simultaneously exploit their user’s data.
So let me explain something that seems obvious to me but which eludes most of the tech luminaries. No, you can’t grow indefinitely. No system in nature is infinite. No ecosystem can be exploited forever. The fallacy that we can push the boundaries as much as we want has no real correspondence with the world.
If anything, nature is showing quite the opposite. Overfishing is decimating the oceans. Pesticides and agrochemicals are reducing the yield, instead of increasing them. Our need for natural resources is devastating to our environment, and with them, our means to survive.
Online Advertising
Despite all this, there is a prevalent view in the technology industry that we can scale our products infinitely. But we can’t. A good example is the rise of online advertising.
Online ads have become the linchpin of nearly every social network in existence. From Facebook to Twitter, all the way to the Chinese giant WeChat (Tencent).
Advertising models allow companies to offer their products for free while monetizing their user’s attention. Providing a product for free ensures rapid growth (assuming good market-fit) and with it, scale.
There is nothing wrong with this model per se. It works, and if done with care, it benefits both sides. Users enjoy a free product, while also getting exposed to products or services they might like. The better targeted the ads are, the less it feels like advertising and turns into useful information. I’m personally impressed with some of the ads on Instagram, as they nail my preferences.
There is one small catch though. To attract big advertisers to any platform, there has to be an excellent way to target them. Brands don’t pay for mass advertising, but to be able to do micro-targeting. Therein lies the problem. Micro-targeting is subjected to knowing a big deal about your users.
The Data Curse
As mentioned before, advertisers look for three things on any platform. To advertise where everyone is, to be able to pinpoint specific segments from the crowd, and to ensure these people see their ads. Scale, segmentation, and attention.
The power of a network, thus, resides in the convergence of these three variables. The total number of active users, how much information the platform knows about them, and how engaged are they.
The more meaningful these metrics are, the more advertising revenue they generate. For most online properties, digital advertising prices are set by brands, and not the platform. The bargaining power sits with the advertisers. However, if the platform is big and sophisticated enough, such power starts to shift to the social network thus achieving pricing power.
“2017 was even more interesting: the company said it would stop increasing ad load in the News Feed, which is why impressions fell, but the price-per-ad increased in response. This demonstration of pricing power is as clear an indication as you can get that Facebook’s News Feed ad was highly differentiated.”
“Facebook’s Story Problem — and Opportunity.” Ben Thompson, Stratechery, Aug 2018.
For a platform to increase all three things, it needs a common thread, data. To attract users, customers need to find value in it. The value of most social networks is grounded in User Generated Content (UGC). Think of Instagram, Youtube, Twitch, etc.
On top of generating a steady stream of data, the network needs to learn all they can about their users. The first step is having the user fill in information about themselves (i.e., birthday, degree, favorite music, etc.). As platforms increase in sophistication, they’ll also be able to analyze the behavioral patterns of their users. The can learn who do you interact with, what type of content you enjoy, when do you take breaks, etc.
Why it matters: Most people only focus on the quantitative face of data. The more data, the more evil the social network is. In truth, the worrisome aspect isn’t how much data they have, but what two unrelated data points say about you. The segmentation magic happens when the platform can ascertain, with a fair degree of probability, personal trends that can’t be determined with a single fact but in combination with different behaviors learned from various sites. You might never say you’re gay, but the system will know. You might never say you’re about to break up with your partner, but the system will know.
The wider the contact surface between the platform and the user, the more the models can infer. The unification of the backend of Messenger, WhatsApp and Instagram should give everyone pause. Connecting all three services allows Facebook to track user behavior during 80% or more of their Internet time.
Last but not least, social networks need to inject advertising in places they know their users will see them. Such a design move is always a conundrum. If users feel advertising is getting in the way of functionality, they’ll stop using it (except if there is a monopoly). Determining where, when and how frequent users consume ads, also requires data from the behavioral analysis.
The big picture: The use of advertising as the core revenue model of a company ensures, by necessity, a data gathering operation. The more data the platform has, the more efficient the ads. However, as the platform keeps collecting data, the user’s privacy starts to erode. There is a point in which the invasion of privacy is so extensive, that the benefits of the network get overshadowed by it.
Model Limitations
In their race to global dominion, many of the technology companies are ignoring such point of equilibrium between privacy and revenues. Facebook has long passed that point. They keep quibbling in the mud, but fail to grasp that the issue isn’t one of product design, but of their business model.
Other companies are following suit. The need for data in advertising models is but one example. With the rise of Deep Learning and new AI methods, the need for datasets is more significant than before.
All of these models have a point of equilibrium; let’s call it the Functionality-Privacy point (FP). If companies push past this point, users will suffer. The loss of privacy isn’t just a moral matter, but a mental health one. Coercing users into engagement submission will break people’s mental and life balance.
Go Deeper: If the use of data for advertising models is terrible, the aggressive use of data in AI models is exponentially worse. When companies rest their product functionality in user data, they should think long and hard about the limitations of such a model. When every system we engage with is requesting, measuring or siphoning behavioral data without consent, users will rebel. There is a need to keep the Functionality-Privacy balance in every company we build.
Data shouldn’t be the economic linchpin of a company. Most AI systems require data, and that’s ok to a certain extent. Nonetheless, the primary revenue model of a company shouldn’t rest on data gathering. If data is at the heart of the model, the company will most surely, invert their priorities, from serving users to helping the capital markets.
In a world where most organizations are heading that way, running a balanced operation will become a value and not a drawback. It’s not far fetched to predict that a whole new crop of companies will make that balance their mighty sword. Apple is already capitalizing on it, but time will say if their underlying model for the next generation of products is pushed beyond the FP limit.
If you like this article, please share it, and invite others to follow the newsletter, it really helps us grow!