Lessons From the Facebook Fiasco
The funniest and the most ironic part of this Facebook fiasco is that target marketing based on data actually worked. If it didn’t, no one would care about some data geeks hoarding data somewhere in the cloud. And any reader of a magazine called “Target Marketing” shouldn’t be surprised by any of this.
Why should it be surprising that, with all of those behavioral and demographic data that Facebook collected, some analytics company could actually figure out who is more likely to vote for Trump? A mediocre-level analyst can do that with data far less immediate and complete than what were used in actuality.
In fact, the whole basis of the Facebook business model is that:
- It is the largest billboard in the world; and
- It can effectively deliver customized messages to each individual.
Facebook has been doing this kind of targeting with or without any third-party analytics vendor such as Cambridge Analytica. The fact that Facebook is expert in grabbing people’s attention (or their eyeballs) for elongated time makes it ever more powerful and effective than third parties; hence, all of the hoopla in the media.
The scarier part (to me, at least) is that a senator actually had to ask Mr. Zuckerberg about his company’s business model, to which he simply answered it shows targeted ads to its users. Seriously, how these fine gentlemen on Capital Hill will regulate any information business is beyond me. Maybe these senators should delegate that work to their grandchildren.
Because I — like readers of this fine publication — do target marketing for living, I wouldn’t blame Facebook for doing things that I would have done myself. And yet, this recent revelation of Facebook’s business with Cambridge Analytica doesn’t sit right with even the most avid practitioners of target marketing. It is not just the fact that it actually affected the presidential election, which is a personal matter for many. Even without any political ramification — which certainly is acting as a magnifying glass in this instance — something certainly went wrong here. Basically, sensitive data were mishandled, and there is no sign of data governance when it comes to sharing data with third-party companies. Considering the influential power of Facebook, no wonder everyone is talking about it as a scandal.
But here, let’s dig into this from a data player’s point of view.
PII – Handle With Care
Facebook is sitting on an amazing amount of personal data, as the users voluntarily share them. Just the profile page alone is a goldmine. Added to that, Facebook has data regarding what the users clicked, liked, shared, reviewed and bought. We are talking about all three major elements of data in target marketing — demographic, behavioral and attitudinal data (refer to “Big Data Must Get Smaller”) all in one place, with an amazing depth in each area. In the business of analytics, that is like having some kind of superpower.
In the early days of Facebook, I voluntarily shared the details of personal information, just to see how good the targeting would be. For the record, to this date, I’m not totally blown away by its targeting accuracy.
Sure, I put down my hobby as guitar playing, and Facebook would show me more guitar stuff. Anyone should be able to do that with such explicit data. But when I go to my Facebook wall, it still feels like I’m looking at a billboard, not a series of targeted messages. Especially the sponsored ads on the right side of the screen. Facebook has been showing me some SUVs (I’m not an SUV guy) and men’s apparel commercials (sure, I’m a man) for days. Meh. That looks like Facebook is just selling regular banner ads, targeted with some rudimentary selection logic using basic data.
Even during the last election cycle, with all of those hyper-political memes and reactions to them floating around, I’ve seen some “totally off the mark” political ads on my wall. What a shame, with all of that rich data in its hands.
So this Cambridge Analytica comes along, and promises the kinds of things that Facebook is supposed to be able to do with all its data for a bunch of politicians. And some bit the bait and paid a good amount of money for the vendor’s services.
Now, this third-party company got to have access to the Facebook user data or a pathway to collect more data on Facebook users. For that matter, any app that runs on Facebook — such as innocuous-looking personality tests — is specifically designed to harvest data. We all do it because it is fun to play, and the cost seems low. Not to offend anyone, all the apps normally require is signing on using Facebook ID. But that is all they really need.
Things get a little tricky there, though; so, who owns the data? The user, Facebook or the third-party vendor? It really depends on that user agreement that 99.99% of users didn’t bother to read. If I must provide an answer as a professional data player, I’d say “all three,” because data collection and refinement warrants some value. Saying a user has the sole ownership of data is like saying that a rice farmer has a right to every piece of sushi sold in restaurants indefinitely.
People who do target marketing for living, in this case or in general, are less scared of such data sharing. What would be the worst thing that could happen? That I get to see an ad of a political candidate who I can’t stand on my Facebook wall? That is, however, if the data are used for general targeting purposes only. Data breaches are indeed scary, because pretty much everything that you put in and you did are linked to your personally identifiable information (PII).
We know that no reputable data player would look up one person at a time by name and see what she is up to. In this case, the operative word is “reputable.” Even the folks who gave permission to Facebook to collect and use data wouldn’t agree that that third-party vendor was indeed reputable. Figuring that out is assumed to be the duty of Facebook, and it is sad that even it didn’t seem to know.
Facebook did not have a good handle on “who gets to use what data.” That is the most unsettling part for people who deal with data for living. Facebook is not exactly dealing with credit card or medical data there, but the sheer volume of data makes the matter as serious. Would it be harsh if I say that Facebook, in pursuit of increasing its revenue and shareholder value, just went for things that it shouldn’t have gone for? Where is the governance? All this mess, for what? Some “semi-accurate” targeting? (I’d love to see some reports on the backend.)
Data-Mining Friends, Too?
It is one thing that Facebook or its partners used data that I shared on Facebook in the form of profile and interests. It is quite another if some shady vendor downloaded the entire list of my friends and call it “its” data source.
That is just a sleazy practice. I’m pretty sure that I did not give permission to share the entire list of my friends with a company that conducts some goofy political spectrum test. No one would put that in an agreement in case just “1” person reads it. Because they themselves would know that that is a sleazy thing to do.
Data players must follow a very simple rule; if you don’t want someone to do certain things with your data, don’t do it yourself (refer to “Don’t Do It Just Because You Can”).
Don’t Be a Data Hoarder
In the business of targeting (or analytics for such targeting), more data don’t always guarantee accuracy. There are all kinds of data out there, and not all data are useful or effective in prediction (refer to “Not All Databases Are Created Equal”). That is why I have been writing repeatedly that one must set the project goal first, not just before some elaborate analytical exercise, but even before data collection.
Mindless hoarding often gets the collector in trouble like we are seeing here. Sometimes “more” data increase only trouble, not the targeting accuracy. Yes, the databases must be broad, accurate, recent and consistent to be useful. But too many data players became too greedy, and there are consequences of being greedy.
Maybe the notion of Big Data gave a wrong impression that big is always good. There are costs involved in dealing with really large data, and another lesson that we must learn here is that a collector can make people mad if “they” think that he is going after too much data.
If the goal is to obtain a “reasonable” level of accuracy in targeting, no, you don’t have to have every piece of data about the target. No analyst likes missing values, but there will be no complete database now or in the future, anyway. A job of analysts is to make the most of what they get, not asking for the entire universe. So, always consider the cost of hoarding too much information, including the social cost.
If all Cambridge Analytica wanted was to predict who was more likely to vote for Trump last year, there were many safer and simpler ways to go about doing that.
Facebook Is Too Powerful?
The ironic part of it all is that Facebook, thanks to its vast coverage, doesn’t require pinpoint targeting precision, anyway. It’s not like it’s going to spend over $1 per piece in direct mailing. Targeted messages are cheap on that platform, and the risk of being wrong is not that high (relatively speaking).
And it seems like Facebook knows it, too. Based on numerous articles that I read about this incident, its analytics is more about maintaining a captive audience by creating a very addictive platform. If the number of eyeballs and time spent on the page are what they are really pursuing, they seem to be doing a fine job there.
If the goal is about increasing targeting accuracy — while not pissing off a great number of people — then it is obvious that Facebook must tighten up its grip on data governance. Like most other data players like us have been doing all along.
Facebook will remain as a powerful force in the market. With or without precision targeting, it’s the biggest billboard in the world. Let’s just say that I didn’t sell off Facebook stocks because of all of this. But if Facebook really wants to benefit human collectives like it used to say in the beginning, lots of significant changes are warranted. And I think that subject is for other forums, not here.
Stephen H. Yu is a world-class database marketer. He has a proven track record in comprehensive strategic planning and tactical execution, effectively bridging the gap between the marketing and technology world with a balanced view obtained from more than 30 years of experience in best practices of database marketing. Currently, Yu is president and chief consultant at Willow Data Strategy. Previously, he was the head of analytics and insights at eClerx, and VP, Data Strategy & Analytics at Infogroup. Prior to that, Yu was the founding CTO of I-Behavior Inc., which pioneered the use of SKU-level behavioral data. “As a long-time data player with plenty of battle experiences, I would like to share my thoughts and knowledge that I obtained from being a bridge person between the marketing world and the technology world. In the end, data and analytics are just tools for decision-makers; let’s think about what we should be (or shouldn’t be) doing with them first. And the tools must be wielded properly to meet the goals, so let me share some useful tricks in database design, data refinement process and analytics.” Reach him at email@example.com.